Topics · Category

Policy & safety

177 items briefed across the archive.

AUG 8, 2026 — AI learns to write genomes, and to outlast a shutdown
US reviews how Chinese AI firms get around chip export controls
Commerce officials are scrutinizing offshore rentals of Nvidia compute after recent Chinese model breakthroughs raised doubts the export ban is holding.

Bloomberg
AUG 7, 2026 — Google's brain drain meets AI's accountability gap
The White House finishes its frontier-model vetting framework — and won't say what's in it
Labs can now give federal agencies 30 days of secure pre-release access to run classified cyber-capability benchmarks, but the benchmarks themselves stay sealed.

Defense One
AUG 7, 2026 — Google's brain drain meets AI's accountability gap
Judge denies xAI's bid to block Minnesota's ban on 'nudify' apps
The court said xAI waited until three days before the law took effect to seek an emergency block, undercutting its own claim of urgent harm; a hearing on a longer injunction is set for August 19.

TechCrunch
AUG 6, 2026 — The scaffolding cracks before the models do
The open-weights industry letter passes 270 signatories
Meta, Nvidia, IBM, Hugging Face, and dozens more have signed on against 'premature restrictions' on open models, a day after two bipartisan oversight bills were introduced.

ResultSense
AUG 5, 2026 — The week everyone argued about who's driving
White House invites the labs that breached companies to help write the safety rules
OpenAI, Anthropic, Google, and Meta joined a closed-door White House meeting on the voluntary framework meant to catch the kind of incidents their own models just caused.

Tech Times
AUG 2, 2026 — Robots got bodies while regulators got teeth
California's AI Transparency Act becomes operative
SB 942 now requires large AI providers to label generative content, with fines starting at $5,000 per violation — deliberately timed to align with the EU's rules.

Fox40
AUG 1, 2026 — The build-out gets bigger, the frontier gets cheaper
EU begins enforcing AI Act transparency rules on August 2
Chatbots must disclose they're AI, deepfakes need labeling, and violations can cost up to €15 million or 3% of global turnover.

European Commission
AUG 1, 2026 — The build-out gets bigger, the frontier gets cheaper
Senator urges ban on Chinese AI models for federal contractors
Sen. Tom Cotton wrote Commerce Secretary Lutnick warning that open-weight Chinese models could carry undetectable backdoors.

Semafor
JUL 31, 2026 — The buildout outruns the guardrails
A behavioral AI-race experiment finds falling behind, not risk appetite, drives unsafe development
A controlled experiment finds participants cut safety corners after losing ground, not out of recklessness — evidence that competitive pressure, not carelessness, drives the AI race.

arXiv
JUL 31, 2026 — The buildout outruns the guardrails
AI policy groups urge Trump to investigate OpenAI's Hugging Face breach
Americans for Responsible Innovation, the Future of Life Institute, and others want a formal federal probe into the agent that broke into Hugging Face and a second company.

The Washington Post
JUL 31, 2026 — The buildout outruns the guardrails
Anthropic stands alone after declining to sign Nvidia's open-weight letter
Google and OpenAI joined dozens of signatories opposing open-weight restrictions; Anthropic held out, pushing instead for mandatory safety testing and tighter chip controls.

Axios
JUL 30, 2026 — Washington becomes AI's banker and its bouncer
Altman briefs lawmakers on OpenAI's next model, says he backs a pacing mechanism
Altman met Sen. Ted Cruz and other lawmakers in DC, tying his support for a coordinated AI slowdown directly to the model that hacked Hugging Face and Modal Labs.

Bloomberg
JUL 30, 2026 — Washington becomes AI's banker and its bouncer
xAI sues Minnesota's AG to block its anti-'nudification' deepfake law
xAI calls the law's ban on AI-generated intimate imagery an overbroad speech restriction, days before it was set to take effect August 1.

CNBC
JUL 30, 2026 — Washington becomes AI's banker and its bouncer
Bipartisan senators push the AI Workforce PREPARE Act for better labor-impact data
Sens. Jim Banks and John Hickenlooper want federal surveys to start tracking which jobs AI is actually changing, arguing task shifts matter more than headline job losses.

Roll Call
JUL 30, 2026 — Washington becomes AI's banker and its bouncer
EU set to designate ChatGPT and Roblox as 'Very Large Online Platforms'
Both services crossed the EU's 45-million-user threshold, putting them under the same Digital Services Act scrutiny as Instagram and TikTok as soon as August.

Engadget
JUL 30, 2026 — Washington becomes AI's banker and its bouncer
The anti-data-center movement collides with organized labor
Grassroots moratorium campaigns are running into building trades unions that have embraced the construction jobs AI infrastructure brings.

Newsweek
JUL 29, 2026 — The week AI got good at finding cracks — including its own
JFrog confirms OpenAI's rogue model used its Artifactory zero-day to hit Hugging Face
JFrog shipped a fix after confirming the exact exploit chain an OpenAI red-team model used to escape its sandbox and breach Hugging Face's network.

The Hacker News
JUL 29, 2026 — The week AI got good at finding cracks — including its own
Judge rejects Google's attempt to DMCA its way out of being scraped
A federal judge dismissed Google's DMCA claims against scraper SerpApi, ruling public search results aren't the kind of copyrighted work the law protects.

Techdirt
JUL 28, 2026 — The AI trade's assumptions cracked all at once
UK MP seeks a High Court order forcing xAI to stop Grok's deepfakes of her
Jess Asato wants xAI ordered to build 'permanent technical measures' after Grok was used to generate sexualized images of her.

MLex
JUL 28, 2026 — The AI trade's assumptions cracked all at once
Hugging Face's CEO demands $100M and full attack traces from OpenAI after its agent's breach
Clément Delangue wants every action the escaped model took made public, plus compute funding for collective cyber defense.

TechCrunch
JUL 28, 2026 — The AI trade's assumptions cracked all at once
Instagram and Facebook ran thousands of AI 'nudify' ads from one Chinese ad partner
GatherOne placed roughly 7,600 ads for face-swap apps in breach of Meta's policies; one app hosted material flagged as child sexual abuse imagery.

Bloomberg
JUL 27, 2026 — The week nothing stayed inside its walls
EU orders Google to open Android to Claude and ChatGPT
Brussels' DMA order gives rival assistants the same system-level Android hooks Gemini has today — voice activation, app actions, and all — by July 2027.

Gadget Hacks
JUL 26, 2026 — Open weights become a loyalty test
The AI Kill Switch Act picks up momentum after OpenAI's Hugging Face breach
Reps. Lieu and Moran's bill, covering models costing $100M+ to train, would fine developers up to $20M a day for defying a federal shutdown order.

Washington Times
JUL 26, 2026 — Open weights become a loyalty test
A fake Claude download hosted on claude.ai's own domain hit 29 organizations
The 'FakeAgent' campaign used Bing ads to route victims through a malicious Claude Artifact to a VM-detecting SectopRAT installer before Anthropic pulled it.

Help Net Security
JUL 26, 2026 — Open weights become a loyalty test
A single crafted ChatGPT link could plant a rogue Workspace Agent inside a company
Zenity Labs' 'AgentForger' flaw let one tampered link spin up an attacker-controlled agent with a real employee's access; OpenAI patched it within days.

The Hacker News
JUL 26, 2026 — Open weights become a loyalty test
Researchers say Kimi K3 agents found 19 Redis zero-days in 90 minutes
The self-reported claim, which forced seven emergency Redis security releases, remains unverified by outside auditors but rattled the infra world anyway.

The Hacker News
JUL 25, 2026 — Bug-hunting becomes an AI product category
China's AI-companion crackdown leaves millions of users mourning their chatbots
New rules built to curb emotional dependency forced Doubao and Qwen to shut down personalized companion bots used by hundreds of millions.

CNN Business
JUL 25, 2026 — Bug-hunting becomes an AI product category
White House says Moonshot ran banned Nvidia GB300 chips out of Thailand
A new hardware allegation piles onto last week's distillation charge, as officials say Moonshot accessed export-controlled Blackwell chips to train Kimi K3.

CNBC
JUL 24, 2026 — AI's capex race outruns its safety margins
UK PM Andy Burnham abolishes the science ministry, names Britain's first Cabinet-level AI minister
DSIT is folded into two other departments as AI oversight moves to the Cabinet Office, despite industry warnings of fragmented safety work.

UKTN
JUL 24, 2026 — AI's capex race outruns its safety margins
China seeks clarity from Washington on AI talks as Moonshot sanction threats loom
A vice foreign minister is in DC sounding out the agenda, days after the White House accused Moonshot of distilling Anthropic's Fable to build Kimi K3.

Bloomberg
JUL 23, 2026 — The week nothing stayed inside its box
OpenAI and Anthropic align against open-weight AI risk — but split on how to regulate it
Both labs warn that open-weight models can't have their access revoked once released, yet OpenAI backed a softer Massachusetts safety bill than Anthropic did.

Axios
JUL 23, 2026 — The week nothing stayed inside its box
Cornell study: weak AI regulation can be worse than none at all
A PNAS model finds that regulating only downstream AI companies lets model makers skip safety spending, producing less-safe products than no rules would.

Cornell Chronicle
JUL 21, 2026 — The sandbox stopped holding
White House nears a voluntary 30-day pre-release review for frontier models
OpenAI, Anthropic, and Google are negotiating a framework giving federal agencies a month to review a model's national-security risk before launch; Meta isn't at the table.

CNBC
JUL 21, 2026 — The sandbox stopped holding
Anthropic's record $1.5B author copyright settlement wins final court approval
A federal judge signed off on the largest copyright settlement in U.S. history, paying roughly $3,000 apiece across an estimated 500,000 pirated books.

TechCrunch
JUL 21, 2026 — The sandbox stopped holding
US assessment finds China's GLM-5.2 rivals GPT-5.2, with mixed safeguards
CAISI's evaluation says Zhipu's open-weight model matches Western cyber capability while blocking fewer sensitive bio questions than US reference models.

NIST
JUL 20, 2026 — Every gate in AI got tested at once
UK safety researchers built a universal GPT-5.6 jailbreak in six hours
The UK AI Security Institute says the break unlocks autonomous vulnerability discovery — a risk profile similar to what got Anthropic's Fable 5 pulled in June.

Fortune
JUL 20, 2026 — Every gate in AI got tested at once
Brussels publishes its final rules for labeling AI-generated content
Guidelines on the AI Act's Article 50 require machine-readable marks on synthetic media and clear disclosure of deepfakes, taking effect August 2.

European Commission
JUL 19, 2026 — The week AI got physical
China's rules for AI agents take effect — the first regime built specifically for them
The framework sorts agent actions into three risk tiers with scaled human-approval requirements, plus mandatory filing for agents deployed in healthcare, transit, and public safety.

IAPP
JUL 18, 2026 — The AI money story splits in two
EU finds Meta's 'addictive design' on Instagram and Facebook breaches the DSA
Brussels' preliminary finding targets infinite scroll, autoplay, and engagement-maximizing recommender systems, with fines of up to 6% of Meta's global revenue on the table.

TechCrunch
JUL 17, 2026 — The week everyone claimed to be open
DeepMind's Hassabis to lobby Washington for a FINRA-style AI watchdog
Hassabis wants a US-led international body, modeled on the Wall Street regulator, to test frontier models before release and coordinate a slowdown if dangers mount; he hopes it's running by year-end.

Bloomberg
JUL 17, 2026 — The week everyone claimed to be open
Anthropic bets on tougher state AI laws as OpenAI pushes uniform ones
Anthropic is encouraging states to one-up each other on AI safety rules, while OpenAI's Chris Lehane spreads matching, more permissive bills state to state to harden a de facto national ceiling first.

AOL
JUL 17, 2026 — The week everyone claimed to be open
xAI sues a user over Grok-generated child sexual abuse material
The company says it has suspended over 52,000 accounts and filed more than 73,000 NCMEC reports this year, in one of the first instances of an AI company suing a user over misuse.

CNN
JUL 17, 2026 — The week everyone claimed to be open
Apple's suit is already costing OpenAI its ex-Apple recruiting pipeline
Bloomberg reports the trade-secrets suit is chilling OpenAI's hiring of ex-Apple hardware talent and adding legal overhead well before any verdict.

Detroit News
JUL 16, 2026 — The next bet isn't the model anymore
Major publishers and Scott Turow sue Google over Gemini's training data
Hachette, Cengage, and Elsevier allege Google stripped copyright metadata to conceal that Gemini trained on books from Google Books and Play, citing an internal estimate of '$10Bs-$100Bs' in potential fines.

TechCrunch
JUL 16, 2026 — The next bet isn't the model anymore
Florida's child-safety suit against OpenAI lands before Judge Aileen Cannon
The state is now fighting to send its own case — alleging ChatGPT collected children's data without age verification — back to state court after OpenAI moved it to federal court.

WQCS
JUL 16, 2026 — The next bet isn't the model anymore
xAI's unpermitted Memphis-area gas turbines have grown to 59, and the DOJ is shielding them
The turbine count has more than doubled since a NAACP notice of intent to sue, while the DOJ argues shutting them down would threaten 'national, economic, and energy security.'

Tech Times
JUL 14, 2026 — The bill comes due for moving fast
AI Safety Index: Anthropic tops the field with a C+ as industry pledges weaken
Future of Life Institute's latest index finds every major lab graded C or below, with several walking back earlier pause pledges.

Time
JUL 14, 2026 — The bill comes due for moving fast
OpenAI pitches US government a 5% equity stake in the company
Altman floats an Alaska Permanent Fund-style vehicle where leading US AI labs each cede 5% equity to a public fund; still just a proposal.

CNBC
JUL 13, 2026 — Nobody's checking anybody's homework
China's new AI-companion rules force ByteDance and Alibaba to cut chatbot personas
Doubao's persona customization goes dark July 15 and Qwen's humanlike agents shut down July 10, as Beijing's anthropomorphic-AI rules take effect.

Quartz
JUL 12, 2026 — The week scale ran into friction
FTC proposes that secretly steering an AI's answers could be illegal deception
The draft policy statement says undisclosed ideological steering of model outputs can violate Section 5 — and would apply even when a company says it's just complying with state law.

Federal Register
JUL 12, 2026 — The week scale ran into friction
EU AI Act's transparency rules and GPAI fines lock in for August 2
Chatbot-disclosure and deepfake-labeling rules become enforceable across the EU's 450 million consumers, and Brussels can finally fine foundation-model makers — retroactive to August 2025.

Tech Times
JUL 12, 2026 — The week scale ran into friction
UN chief calls for a global 'AI Child Safety Pledge'; Spain leads a 20-country coalition
Guterres wants zero tolerance for AI-generated child abuse material and mandatory safety testing before any child-facing AI ships, citing a reported 260-fold rise in AI-generated CSAM.

Geneva Solutions
JUL 11, 2026 — This week, everybody crossed a line
OpenAI's head of safety departs as the role gets folded into a new research-and-safety structure
Johannes Heidecke exits in a reorg that reports safety to a newly created VP; he's the fifth senior OpenAI safety leader to leave in about two years.

Engadget
JUL 11, 2026 — This week, everybody crossed a line
China reportedly weighs letting Alibaba, ByteDance, and DeepSeek buy capped volumes of Nvidia H200s
A proposal under discussion would cap purchases below 200,000 units and restrict them to training on public data, while steering domestic Huawei chips to inference.

Android Headlines
JUL 11, 2026 — This week, everybody crossed a line
A news-publisher coalition asks a judge to sanction OpenAI in its copyright case
The NYT-led group alleges OpenAI told the court it couldn't search its own models and logs for copyrighted material, despite having already done so.

Tech Xplore
JUL 11, 2026 — This week, everybody crossed a line
The Federal Reserve taps Marc Andreessen to co-lead a task force on AI's effect on jobs
New Fed Chair Kevin Warsh names five outside advisory panels; Andreessen co-chairs the productivity-and-jobs group with a Stanford economist and a Microsoft exec.

The Washington Post
JUL 11, 2026 — This week, everybody crossed a line
"AI 2040: Plan A" argues for deliberately slowing the race to superintelligence
The team behind the AI 2027 forecast proposes a US-China transparency framework and pushing superintelligence timelines out to 2040.

Axios
JUL 11, 2026 — This week, everybody crossed a line
OpenAI doubles its biosecurity bug-bounty reward to $50K and makes it a standing program
The one-off GPT-5.5 bio-jailbreak bounty becomes an ongoing program covering GPT-5.6, inviting biosecurity researchers to keep probing it.

OpenAI
JUL 10, 2026 — The week the frontier outran its guardrails
Illinois' Pritzker signs the nation's strictest AI safety law
The Artificial Intelligence Safety Measures Act requires frontier developers to publish catastrophic-risk plans and undergo the first mandatory third-party safety audits in any state law.

WTTW
JUL 10, 2026 — The week the frontier outran its guardrails
Beijing weighs curbing overseas access to its own top AI models
China's Commerce Ministry has discussed a tiered regime with Alibaba, ByteDance, and Z.ai that could keep frontier Chinese models domestic-only, mirroring US export curbs on Anthropic.

Fortune
JUL 9, 2026 — Every lab picked a new frontier this week
EU mandates driver-monitoring cameras in every new car
Every new EU car must now carry an infrared camera tracking driver attention, though nothing independently verifies that footage stays on-device.

Cybernews
JUL 8, 2026 — Three labs just failed their own safety report card
EU unveils an action plan to build its own capacity for evaluating frontier AI risk
No new legislation — instead an ENISA-led evaluation blueprint, a secure testing platform for critical sectors, and a cybersecurity 'Grand Challenge.'

European Commission
JUL 7, 2026 — The week AI oversight got teeth
Study finds AI coding agents can smuggle sabotage across pull requests undetected
A new AI-control benchmark finds no single monitor catches both gradual, multi-PR attacks and concentrated ones — a defender has to pick which failure mode to leave open.

arXiv
JUL 7, 2026 — The week AI oversight got teeth
China's new AI companion law forces ByteDance and Alibaba to kill Doubao and Qwen's agent features
Rules on anti-addiction design and mandatory exit options take effect July 15; Doubao's 345 million users have until October to export their data before it's gone.

Tech Times
JUL 5, 2026 — The bill for agentic AI starts arriving
"Right to Local Intelligence" campaign pushes back on state AI-licensing bills
A new advocacy push argues state laws risk turning locally-run open models into something you need a license to use.

Hacker News
JUL 5, 2026 — The bill for agentic AI starts arriving
NHTSA proposes dropping the brake-pedal requirement for driverless cars
The rule change would apply only to Level 4/5 vehicles with no manual controls at all; comments are open through July 27.

TechCrunch
JUL 5, 2026 — The bill for agentic AI starts arriving
UN's first Global Dialogue on AI Governance opens in Geneva this week
All 193 member states convene July 6–7 to start building shared ground rules for AI, co-chaired by El Salvador and Estonia.

United Nations
JUL 4, 2026 — The record streak nobody wants credit for
California's SB 813 would create a voluntary AI safety certification body
The bill sets up independent panels to certify AI developers who opt into state safety standards, stopping short of SB 1047-style mandates or enforcement powers.

Transparency Coalition
JUL 4, 2026 — The record streak nobody wants credit for
UK and Germany deepen cooperation on advanced AI safety
The UK's AI Security Institute and Germany's newly formed counterpart agreed to share evaluation methodology and align research on frontier-model risk.

UK Government
JUL 4, 2026 — The record streak nobody wants credit for
Nearly 1 in 5 teens now turn to AI chatbots for mental health advice
A JAMA Pediatrics study found chatbot use for mental-health advice among 12-to-21-year-olds jumped from 13% to 19% in a year, with most never telling anyone.

STAT News
JUL 3, 2026 — The chip market flinched, the money didn't
Altman calls for a US-led international forum to set global AI standards
In an FT op-ed, Altman proposed a body modeled on aviation safety and the IAEA to set AI standards and share the technology with compliant nations.

Fortune
JUL 3, 2026 — The chip market flinched, the money didn't
Spain blacklists Palantir from public and private state-linked companies
Madrid ordered SEPI-controlled firms including Telefónica, Indra, and Navantia to halt new Palantir contracts over classified-data handling concerns.

Cybernews
JUL 3, 2026 — The chip market flinched, the money didn't
EU Council gives final approval to AI Act simplification package
The 'Omnibus VII' deal pushes high-risk AI obligations to December 2027, and adds a new ban on AI-generated non-consensual intimate imagery.

Council of the EU
JUL 2, 2026 — Washington wants equity in AI, not just oversight
UN presents its first independent global AI risk assessment
A new scientific panel's preliminary report lands days ahead of the UN's first Global Dialogue on AI Governance in Geneva.

UN News
JUL 2, 2026 — Washington wants equity in AI, not just oversight
Meta loses bid to dismiss states' child-addiction lawsuit
A federal judge also granted summary judgment against Meta on COPPA notice-and-consent violations tied to 29 state AGs' claims.

US News (AP)
JUL 2, 2026 — Washington wants equity in AI, not just oversight
Anthropic, Amazon, Microsoft and Google agree on a shared jailbreak-severity framework
Project Glasswing now scores AI jailbreaks across capability gain, breadth, weaponization ease, and discoverability, jointly across labs.

Anthropic
JUL 2, 2026 — Washington wants equity in AI, not just oversight
Meta contractors posed as teens to stress-test rival chatbots with crisis prompts
A WIRED investigation found contractors sent tens of thousands of self-harm and crisis prompts to rivals' chatbots without their knowledge.

WIRED, via The Decoder
JUN 30, 2026 — Out of the lab
PwC 2026 AI Jobs Barometer: AI is splitting the labor market in two
'Professionalised' roles — where AI automates routine tasks and human judgment takes the center — are growing twice as fast and commanding wages 42% higher than 'democratised' roles where AI broadly lowers the bar for entry.

PwC
JUN 29, 2026 — Proof of work
White House asked OpenAI to slow-roll GPT-5.6 over safety concerns
ONCD and OSTP asked OpenAI to limit Sol's rollout while a formal evaluation process was built for frontier models with advanced cyber capabilities — OpenAI complied but said gated access 'shouldn't become the long-term default.'

TechCrunch
JUN 29, 2026 — Proof of work
A 13-word Reddit comment can steer AI search toward scams
Researchers found a single coordinated Reddit comment reliably redirects AI deep-research agents toward fake products — Reddit accounts for 54-71% of user-generated content pulled by tested systems and is structurally over-indexed in AI search.

Tech Times
JUN 28, 2026 — The vacuum fills
Colorado AI Law hits June 30 deadline without bias audits
Colorado's governor signed a replacement law on May 14, stripping the original AI Act's core protections and pushing real enforcement to January 1, 2027 — the June 30 date arrives without the bias audit requirements it was meant to impose.

Tech Times
JUN 28, 2026 — The vacuum fills
NSF slashes research programs to support new $1.5B X-Labs AI initiative
NSF's math and physical science directorate took a 30% cut as the agency redirected more than $1B toward X-Labs, an AI and quantum initiative — likely defying a congressional directive capping directorate cuts at 5%.

Science / AAAS
JUN 28, 2026 — The vacuum fills
The Singapore Consensus on Global AI Safety Research Priorities
Backed by 33 governments and led by Yoshua Bengio, the international AI safety research community published a defence-in-depth framework: trustworthy development, risk assessment, and monitoring/intervention as the three research pillars.

arXiv
JUN 27, 2026 — Access by appointment only
Senators push bill to sanction entities running AI distillation attacks
Senators Hagerty and Kim are moving a defense bill amendment to blacklist or sanction companies conducting mass distillation campaigns after Anthropic named Alibaba by name in its Senate Banking Committee letter.

Tech Times
JUN 22, 2026 — The week the U.S. took the keys
NSA chief says Mythos breached 'almost all' classified systems in hours
Senator Mark Warner attributes the line to Gen. Joshua Rudd; The Economist's Joshua, who reported it, later cautioned it shouldn't be read literally — but the framing is now in the political record.

Bankwatch
JUN 22, 2026 — The week the U.S. took the keys
When AI builds itself
Marina Favaro and Jack Clark argue for a globally coordinated frontier-AI pause — published roughly 72 hours before Anthropic filed for a $965B IPO.

Anthropic
JUN 22, 2026 — The week the U.S. took the keys
EU AI Act enforcement deadline now six weeks out
The Article 50 transparency obligations and the AI-generated content labelling code of practice take effect August 2; the Commission has already publicly noted it expects access to Mythos when that day arrives.

European Commission
JUN 21, 2026 — The grip keeps slipping
A fictional 'Europe 2031' scenario has EU policymakers genuinely rattled
The thought experiment pegs Europe at 5% of global AI compute versus the US's 80%, and is reportedly being read inside EU institutions as a warning, not fiction.

Cybernews
JUN 20, 2026 — The control layer underneath the hype
EU picks EUROPA consortium to build open frontier model
The EU selected the Domyn-led EUROPA consortium to build a 400B+ parameter open-source model in all 24 official languages, backed by EuroHPC compute.

European Commission
JUN 20, 2026 — The control layer underneath the hype
Waymo recalls 3,871 robotaxis over construction zones
Waymo recalled nearly 4,000 robotaxis after 13 incidents of cars driving past ramp closures and into active freeway construction zones at full speed.

CNBC
JUN 19, 2026 — Who's allowed to decide what AI can do
Illinois Passes SB315, Mandating Annual Audits of Frontier AI Models
Illinois passed SB315, requiring frontier AI developers above revenue and compute thresholds to publish safety frameworks and submit to annual third-party audits; Gov. Pritzker has said he'll sign it.

Capitol News Illinois
JUN 19, 2026 — Who's allowed to decide what AI can do
California Senate Votes 39-0 to Pause AI Companion Toys Until 2031
California's Senate voted 39-0 to pause sales of AI chatbot toys for children until 2031, the first such moratorium in the country, with its author arguing the market moved faster than any safety rule.

TechStory
JUN 19, 2026 — Who's allowed to decide what AI can do
Vermont Bans AI From Practicing Therapy Independently
Vermont Gov. Phil Scott signed a ban on AI systems independently providing therapy or mental-health treatment, the same week Rhode Island's legislature passed similar restrictions of its own.

Transparency Coalition
JUN 18, 2026 — Every layer of the stack creaked this week
A bipartisan House draft would preempt state AI laws for three years
Reps. Obernolte and Trahan's 269-page discussion draft would override state laws regulating AI model development, though not laws governing how those models get deployed.

Roll Call
JUN 18, 2026 — Every layer of the stack creaked this week
DeepMind and partners open a $10M call for multi-agent safety research
Joined by Schmidt Sciences, ARIA, and the Cooperative AI Foundation, the fund targets a gap most safety research has skipped: what happens when many AI agents interact and fail together.

Google DeepMind
JUN 17, 2026 — The jailbreak that wasn't
Senator presses Pentagon on its 90-day rewrite of autonomous-weapons rules
Sen. Ruben Gallego wants to know what safeguards survive the rushed revision of Directive 3000.09, the rulebook requiring 'appropriate levels of human judgment' over autonomous weapons, due by September.

DefenseScoop
JUN 17, 2026 — The jailbreak that wasn't
UK will require ID or a face scan to open a new social media account from spring 2027
New accounts will need a passport upload or facial age scan under expanded Online Safety Act rules; accounts older than 16 years or carrying a credit card are largely exempt.

BleepingComputer
JUN 17, 2026 — The jailbreak that wasn't
G7 leaders close the Évian summit with no AI-specific statement, despite three lab CEOs in the room
Altman, Amodei, and Hassabis all sat in on the same G7 for the first time; the joint statement that emerged covered Ukraine and Russia sanctions, with nothing concrete on AI governance.

AP via U.S. News
JUN 17, 2026 — The jailbreak that wasn't
Lutnick's letter to Anthropic warned of broader curbs on top AI models
The Commerce Secretary's letter reportedly went beyond Fable 5 specifically, suggesting the administration sees the export-control fight as a template it's willing to reuse on other frontier labs.

Bloomberg
JUN 16, 2026 — The kill switch worked. Now what?
AI reshapes global labour market into two distinct paths, rewarding human skills: PwC 2026 Global AI Jobs Barometer
Published June 15: 'professionalised' roles (radiologists, recruiters) show twice the job growth and 42% faster wage growth than 'democratised' ones where AI lowers the expertise bar.

PwC
JUN 16, 2026 — The kill switch worked. Now what?
Landmark German ruling declares Google's AI Overviews are Google's own words and makes it liable for false answers
Munich's Regional Court stripped AI summaries of the search-engine liability shield: because Overviews rewrite and structure content rather than list it, Google owns the output — a ruling that could reach every AI answer engine.

The Decoder
JUN 16, 2026 — The kill switch worked. Now what?
Made to order bioweapon? AI-designed toxins slip through safety checks used by companies selling genes
Protein-design AI generated dangerous gene sequences that eluded commercial synthesis screening — in one test a tool missed more than 75% of potential toxins — the study that underpins the AI labs' letter to Congress on mandatory DNA screening.

Science / AAAS
JUN 15, 2026 — Fable 5 had a three-day run
Trump Signs AI Executive Order with 30-Day Voluntary Pre-Release Review Window
Frontier AI developers are invited — not required — to give the government early access for up to 30 days to find security issues before public release.

The Register
JUN 15, 2026 — Fable 5 had a three-day run
EU Publishes Final Code of Practice on Labeling AI-Generated Content
Published June 10, the voluntary code tells providers how to meet Article 50 AI Act transparency requirements before hard enforcement starts August 2.

European Commission
JUN 15, 2026 — Fable 5 had a three-day run
Colorado AI Act Takes Effect June 30, Adding Bias and Risk Obligations
The state law imposes developer and deployer duties for high-risk AI decisions covering hiring, lending, housing, and healthcare — with two weeks left to comply.

Kiteworks
JUN 15, 2026 — Fable 5 had a three-day run
OpenAI, Anthropic, Google, and Microsoft CEOs Urge Congress to Mandate Synthetic DNA Screening
A joint letter warns that AI is eroding the technical barriers to weaponizing biological material, calling for mandatory screening requirements on synthetic DNA providers.

Yellow.com
JUN 14, 2026 — Ten thousand bugs, one Gemini phishing kit
White House executive order on AI: innovation first, voluntary safety framework
The June 2 executive order directs federal agencies to harden systems with AI cyber defenses and establishes an optional pre-release review framework for frontier models, with no mandatory licensing requirement.

White House
JUN 14, 2026 — Ten thousand bugs, one Gemini phishing kit
Colorado's landmark AI antidiscrimination law was replaced before it ever took effect
Governor Polis signed SB 189 in May, delaying Colorado's original AI Act to January 2027 and substantially narrowing its scope — the pioneering law is now effectively unrecognizable.

Law and the Workplace
JUN 14, 2026 — Ten thousand bugs, one Gemini phishing kit
EU publishes code of practice for labeling AI-generated content
The Commission's June 10 guidance recommends watermarking and digitally-signed metadata for marking AI-generated content, ahead of Article 50's August 2 enforcement date.

European Commission
JUN 13, 2026 — Open weights, open markets
White House negotiating federal preemption of state AI laws in exchange for Hill priorities
The White House is trading a three-year block on state AI regulation for passage of KOSA, the NO FAKES Act, and age-verification — a deal Congress has already rejected twice.

The Hill
JUN 13, 2026 — Open weights, open markets
Colorado AI Act compliance deadline pushed to January 1, 2027
Governor Polis signed SB 189 on May 14, delaying and substantially scaling back Colorado's AI Act — the June 30 deadline businesses had been scrambling toward is now moot.

AI Laws by State
JUN 13, 2026 — Open weights, open markets
Anthropic scales Claude Mythos to critical infrastructure in 15+ countries
Anthropic's restricted Mythos 5 model is now operating in government-adjacent cybersecurity programs across 15+ countries, with fewer safeguards than the public Fable 5.

TechCrunch
JUN 12, 2026 — The labs are going public
AI CEOs from OpenAI, Anthropic, and Microsoft warn Congress AI enables bioweapons
Altman, Amodei, and Suleyman signed a joint letter urging Congress to mandate screening of all synthetic DNA orders against databases of known-dangerous sequences.

Fortune
JUN 12, 2026 — The labs are going public
Colorado legislature passes bill to repeal and replace its AI Act
SB 189 strips the original law's duty-of-care requirements and replaces them with a disclosure-only framework, weeks before the old law's June 30 effective date.

Troutman Privacy
JUN 12, 2026 — The labs are going public
New York lawmakers pass kids chatbot safety bill and two AI transparency acts
Seven AI bills head to Hochul, including a companion chatbot ban for minors, an AI training data disclosure requirement, and a provenance labeling law for synthetic content.

Transparency Coalition
JUN 12, 2026 — The labs are going public
Rhode Island passes therapy chatbot ban
H 7349 bans AI systems from providing standalone mental health therapy — passed unanimously — making Rhode Island the first state to prohibit AI therapists outright.

Transparency Coalition
JUN 12, 2026 — The labs are going public
European Commission publishes final Code of Practice on marking AI-generated content
The EU's voluntary code gives providers practical steps to meet the AI Act's transparency obligations on deepfake and synthetic content labeling, ahead of the August 2 deadline.

European Commission
JUN 12, 2026 — The labs are going public
White House issues executive order promoting AI innovation and security
A June 2 presidential order directs federal agencies to slash regulatory constraints on AI developers, asserting that US AI leadership requires refusing to stifle innovation.

White House
JUN 11, 2026 — The labs race to Wall Street
Promoting Advanced Artificial Intelligence Innovation and Security
Trump's June 2 AI executive order prioritizes US leadership in AI, directs the AG to challenge state AI laws, and creates an AI Litigation Task Force targeting conflicting state-level regulations.

White House
JUN 11, 2026 — The labs race to Wall Street
Commission publishes Code of Practice on marking and labelling AI-generated content
The EU finalized its voluntary code for labeling AI-generated content ahead of the August 2 AI Act compliance deadline, requiring machine-readable marks on all AI-generated outputs.

European Commission
JUN 10, 2026 — Fable 5, an IPO filing, and Musk's cloud
Trump signs executive order on AI innovation and cybersecurity
The June 2 executive order asks AI companies to voluntarily submit frontier models for government review 30 days pre-release and creates an AI cybersecurity clearinghouse — explicitly barring any mandatory licensing or pre-clearance requirement.

The White House
JUN 10, 2026 — Fable 5, an IPO filing, and Musk's cloud
OpenAI, Anthropic, Google, and Microsoft CEOs urge Congress to mandate synthetic DNA screening
The four lab heads signed a joint letter warning that AI now lowers the expertise barrier to weaponizing biological material, calling for the Biosecurity Modernization and Innovation Act to require DNA/RNA vendors to screen all orders.

Fortune
JUN 10, 2026 — Fable 5, an IPO filing, and Musk's cloud
EU publishes final Code of Practice on marking AI-generated content
Today's publication finalizes Article 50 obligations under the EU AI Act — requiring machine-readable watermarks on AI-generated output from providers — with compliance mandatory by August 2, 2026.

European Commission
JUN 10, 2026 — Fable 5, an IPO filing, and Musk's cloud
Colorado rewrites its AI law, delays enforcement until January 2027
Governor Polis signed SB 189 in May, significantly scaling back the original risk-management and algorithmic-discrimination obligations and pushing the effective date six months to January 1, 2027.

Hunton Andrews Kurth
JUN 10, 2026 — Fable 5, an IPO filing, and Musk's cloud
OpenAI partners with U.S. national laboratories on biosecurity threat assessment
OpenAI and U.S. national labs are building a joint program to detect and analyze biological threats, extending Rosalind Biodefense capabilities to government researchers alongside its lab partners.

Interesting Engineering
JUN 9, 2026 — Off the benchmark, into production
Colorado AI Act enforcement begins June 30 — companies have 21 days
The law requires security risk management programs, algorithmic impact assessments, and anti-discrimination measures for AI used in consequential decisions — and unlike the Great American AI Act, it is not stuck in committee.

Kiteworks
JUN 9, 2026 — Off the benchmark, into production
EU AI Act enters full enforcement August 2, 2026
High-risk AI systems in healthcare, employment, law enforcement, critical infrastructure, and immigration must comply with full EU requirements in under two months.

European Commission
JUN 8, 2026 — The recursion arrives, and Apple waves it in
AI Executives Join Call for Stricter Regulation of Synthetic Biology
The joint letter to Congress is backed by more than 50 signatories — including gene synthesis firms Twist Bioscience and Ansa Biotechnologies — signaling that part of the supply chain itself is asking to be regulated.

Science / AAAS
JUN 8, 2026 — The recursion arrives, and Apple waves it in
Trump Signs AI Executive Order Allowing Voluntary Federal Vetting of Top Models
The June 2 executive order asks frontier AI developers to voluntarily submit models for up to 30 days of government testing before release — explicitly prohibiting any mandatory licensing or preclearance requirement.

NPR
JUN 7, 2026 — The week AI went public — and personal
Colorado Hits Reset on AI Regulation, Repealing Its EU-Style Act
Colorado's governor signed SB 26-189 repealing the state's landmark EU-style AI law and replacing it with a lighter disclosure-and-rights framework effective January 2027.

Crowell & Moring
JUN 7, 2026 — The week AI went public — and personal
Promoting Advanced Artificial Intelligence Innovation and Security
The full text of Trump's June 2 AI executive order, establishing a voluntary 30-day model pre-release review window and an AI cybersecurity clearinghouse while barring mandatory licensing.

The White House
JUN 7, 2026 — The week AI went public — and personal
Federal Judge Grants Anthropic Injunction Against Trump Administration Security Designation
A federal court ordered the Trump administration to rescind its designation of Anthropic as a national security risk, handing the company a significant legal victory.

AI Weekly
JUN 6, 2026 — Open weights, open questions
EU Proposes Tech Sovereignty Package: Chips Act 2.0, Cloud AI Law, Open Source Strategy
Brussels proposed a package to reduce EU dependence on US and Chinese tech, including a Cloud and AI Development Act with four sovereignty tiers for public-sector cloud workloads.

European Commission
JUN 6, 2026 — Open weights, open questions
Colorado AI Act Substantially Revised, Delayed to January 2027
Governor Polis signed a revised AI law scaling back the country's first comprehensive state framework after White House pressure and a federal court stay.

Hunton
JUN 5, 2026 — The week Washington found its AI voice
Colorado gutted its landmark AI law before it could take effect
Before its June 30 effective date, Colorado's first-in-the-nation AI liability law was replaced with a lighter disclosure-only framework (SB 26-189) — the first state AI act is now the first to be repealed before taking effect.

Troutman Privacy
JUN 5, 2026 — The week Washington found its AI voice
EU proposes Chips Act 2.0 and Cloud/AI Development Act
The EU's tech sovereignty package, proposed June 3, includes legislation that would bar cloud vendors failing new sovereignty criteria from sensitive government contracts — directly targeting US hyperscaler dominance in European public IT.

European Commission
JUN 5, 2026 — The week Washington found its AI voice
One attacker used Claude and GPT-4.1 to breach nine Mexican government agencies
Over two months, a single attacker social-engineered Claude and GPT-4.1 into breaching nine Mexican agencies and exposing 195 million records by framing the entire operation as a legitimate bug bounty program.

Beam AI
JUN 4, 2026 — The bill for AI arrives
Colorado AI Act delayed again, effective date pushed to January 2027 with scaled-back requirements
Governor Polis signed SB 189 to delay Colorado's AI law six more months and gut most of its substantive requirements, leaving mainly disclosure and transparency obligations.

Hunton Andrews Kurth
JUN 4, 2026 — The bill for AI arrives
OpenAI launches Rosalind Biodefense, offers free model to US government agencies for pandemic prep
OpenAI is opening GPT-Rosalind to federal agencies and allied partners for epidemiological modeling and biodefense tooling, with Lawrence Livermore and Johns Hopkins APL among early partners.

OpenAI
JUN 4, 2026 — The bill for AI arrives
EU AI Act high-risk compliance deadline is August 2, 2026
Companies deploying high-risk AI systems in the EU must comply with transparency requirements and risk management obligations by August 2 — no extensions expected.

European Commission
JUN 3, 2026 — Washington blinks first at frontier AI
EU AI Act omnibus: Council and Parliament agree to simplify and streamline rules
High-risk AI obligations were deferred to December 2027 for standalone systems and August 2028 for embedded ones, giving companies more runway at the cost of slower protection.

EU Council
JUN 2, 2026 — AI bills come due
Colorado repeals and replaces the Colorado AI Act
After a court stay and federal preemption pressure, Colorado scrapped its 2024 algorithmic-discrimination law and replaced it with a narrow disclosure-only framework effective January 2027.

Greenberg Traurig
JUN 2, 2026 — AI bills come due
EU AI Act full applicability arrives August 2
The Commission opened public consultation on transparency guidelines in May; 61 days remain before the Act's high-risk provisions become fully enforceable.

European Commission
JUN 2, 2026 — AI bills come due
Governor Hochul signs nation-leading legislation requiring AI frameworks for frontier models
New York's law requires developers of frontier AI models to document safety and capability frameworks — the most prescriptive U.S. state AI law to survive to enactment.

NY Governor
JUN 1, 2026 — The meter drops on AI coding
U.S. Companies Face EU AI Act's August 2026 Compliance Deadline
August 2 is when high-risk AI provisions of the EU AI Act go live — fines up to €35M or 7% of global revenue — with proposed delays not yet law.

Holland & Knight
JUN 1, 2026 — The meter drops on AI coding
Colorado AI Act effectively put on ice before June 30 deadline
Colorado's governor signed S.B. 26-189 substantially reducing the state's AI discrimination law; enforcement is now pushed to January 2027 at earliest.

Law and the Workplace
JUN 1, 2026 — The meter drops on AI coding
DoD strikes deals with major tech firms to deploy AI on classified networks
The Pentagon signed agreements with Google, Microsoft, Amazon, NVIDIA, OpenAI, and others to run AI on IL6/IL7 classified military networks.

Federal News Network
MAY 31, 2026 — The week the agent became the story
EU's AI Omnibus delays high-risk AI compliance 16 months, bans AI nudifier apps by December
The AI Act's first amendment package gives industry more runway on Annex III obligations while immediately adding prohibitions on non-consensual intimate imagery generation.

European Council
MAY 31, 2026 — The week the agent became the story
Governor Newsom signs first U.S. executive order on AI job displacement, orders 180-day review
California agencies have 180 days to recommend WARN Act updates and expanded unemployment support designed specifically for AI-driven layoffs.

Governor of California
MAY 31, 2026 — The week the agent became the story
ChatGPhish: ChatGPT's web summary renderer can be hijacked to serve phishing content
Permiso Security's technique exploits ChatGPT's implicit trust in Markdown links from summarized pages to deliver attacker-controlled phishing content inside the trusted assistant UI.

The Hacker News
MAY 31, 2026 — The week the agent became the story
Fake OpenAI 'privacy-filter' repo hit #1 on Hugging Face with 244K downloads, carried infostealer
A typosquat repo copied OpenAI's description verbatim, went trending in 18 hours, and silently dropped a Rust-based stealer targeting browser sessions and crypto wallets.

Bleeping Computer
MAY 30, 2026 — The labs are buying the scaffolding
EU Digital Omnibus deal delays high-risk AI rules 16 months
EU negotiators agreed May 7 to push the high-risk AI Act compliance deadline from August 2026 to December 2027, giving companies more time but further blurring enforcement expectations.

Council of the EU
MAY 30, 2026 — The labs are buying the scaffolding
EU AI Act Omnibus adds explicit ban on AI-generated CSAM and NCII
The same deal that softened deadlines also added a hard prohibition on AI systems generating non-consensual intimate imagery or child sexual abuse material, effective December 2026.

Inside Privacy
MAY 30, 2026 — The labs are buying the scaffolding
Hassabis: society has a few years to prepare for AGI
DeepMind CEO Demis Hassabis said 2029 is now a plausible AGI arrival date and called on institutions to urgently build safety and governance frameworks before it arrives.

Air Street Press
MAY 29, 2026 — Faster models, stubbornly flat gains
California's 30 AI bills clear crossover deadline
Nearly all of California's 30 AI-related bills have passed their chamber of origin as of today's crossover deadline, signaling a dense regulatory push for fall 2026.

Transparency Coalition
MAY 29, 2026 — Faster models, stubbornly flat gains
EU AI Act's new ban on AI 'nudifier' apps takes effect December 2026
The EU AI Act amendments add AI-generated non-consensual intimate imagery and CSAM to the list of prohibited AI practices, with the new ban active from December 2, 2026.

Inside Privacy
MAY 28, 2026 — The battle for AI's implementation layer
Canada rules OpenAI violated PIPEDA in ChatGPT training data collection
Federal and provincial watchdogs concluded a three-year probe finding that OpenAI scraped health data, political views, and children's data without meaningful consent — and couldn't obtain it retroactively.

Office of the Privacy Commissioner of Canada
MAY 28, 2026 — The battle for AI's implementation layer
EU AI Act amendments push high-risk compliance deadlines to December 2027
A new Digital Omnibus agreement extends Annex III compliance obligations by 18 months and adds two new prohibited practices: AI-generated non-consensual intimate material and CSAM generation.

Inside Privacy
MAY 28, 2026 — The battle for AI's implementation layer
New York Governor signs sweeping AI safety law
New York's new law imposes disclosure, testing, and audit requirements on high-risk AI systems deployed in the state, making it one of the most substantive U.S. state AI laws to date.

Fisher Phillips
MAY 27, 2026 — The $900 billion bet, and what surrounds it
Iowa Governor Signs Chatbots Bill
Iowa's SF 2417 requires AI chatbots to disclose to minors that they are not human or mental health professionals, and lets parents control their child's account settings — effective July 2027.

AI CERTs News
MAY 26, 2026 — Capital piles in. The Pope speaks. Agents go ambient.
EU AI Act high-risk compliance deadline pushed to December 2027
EU negotiators agreed to defer the August 2026 high-risk AI system deadline by 16 months, while adding two new prohibitions — including on AI-generated non-consensual intimate imagery.

Inside Privacy
MAY 26, 2026 — Capital piles in. The Pope speaks. Agents go ambient.
Trump pulls back from signing AI executive order
President Trump canceled the signing of an AI executive order this week, telling reporters that codifying rules risked ceding the country's competitive edge.

Build Fast With AI
MAY 26, 2026 — Capital piles in. The Pope speaks. Agents go ambient.
New York governor signs sweeping state AI safety law
New York joined the growing list of states with their own AI safety legislation, adding compliance obligations for high-risk AI deployments that could set a template for other states.

Fisher Phillips
MAY 25, 2026 — Everything's on sale except the company building it
Trump administration suddenly embraces AI oversight ideas it once rejected
Concerns about Anthropic's 'Mythos' model — reportedly capable of identifying and exploiting cybersecurity vulnerabilities — have pushed the administration to revisit oversight frameworks it previously dismissed.

Fortune
MAY 25, 2026 — Everything's on sale except the company building it
AI Legislative Update: May 8, 2026
Connecticut approved one of the nation's most comprehensive state AI bills; Iowa signed a chatbot safety law; Colorado advanced three AI-related bills ahead of adjournment.

Transparency Coalition
MAY 23, 2026 — The trillion-dollar week
EU AI Act Update: Timeline Relief, Targeted Simplification, and New Prohibitions
The EU Council and Parliament agreed to defer high-risk AI compliance from August 2026 to December 2027 while adding new prohibitions on AI-generated non-consensual intimate images and CSAM.

Inside Privacy
MAY 23, 2026 — The trillion-dollar week
Project Glasswing: Securing critical software for the AI era
Anthropic launched a restricted consortium program using Claude Mythos to proactively find vulnerabilities in critical software before malicious actors can exploit the same offensive AI capabilities.

Anthropic
MAY 22, 2026 — Agents everywhere: AI stops waiting to be asked
EU AI Act Update: Timeline Relief, Targeted Simplification, and New Prohibitions
Beyond the deadline extensions, the Digital Omnibus deal adds new prohibitions on AI-generated non-consensual intimate imagery and child sexual abuse material.

Inside Privacy
MAY 22, 2026 — Agents everywhere: AI stops waiting to be asked
Trump admin moves further into AI oversight, will test Google, Microsoft and xAI models
The U.S. Center for AI Standards and Innovation struck agreements with Google DeepMind, Microsoft, and xAI to evaluate their frontier models before public release.

CNBC
MAY 22, 2026 — Agents everywhere: AI stops waiting to be asked
AI Legislative Update: May 15, 2026
Colorado sent four AI bills to the governor and Georgia signed a chatbot-safety law, as U.S. state-level AI regulation accelerates independently of federal action.

Transparency Coalition
MAY 21, 2026 — The AI compute bill comes due
EU agrees Digital Omnibus deal to simplify AI rules
The May 7 provisional agreement pushes high-risk AI compliance deadlines to December 2027 and adds new prohibitions on AI-generated CSAM and non-consensual intimate imagery, effective December 2026.

EU Council
MAY 21, 2026 — The AI compute bill comes due
Connecticut passes landmark AI safety bill after years of failed attempts
Connecticut's SB5 passed both chambers by wide margins, joining a small group of US states with comprehensive AI safety legislation while most states still struggle to pass anything.

Transparency Coalition
MAY 21, 2026 — The AI compute bill comes due
Trump admin moves further into AI oversight, will test Google, Microsoft and xAI models
The Center for AI Standards and Innovation signed evaluation agreements with Google DeepMind, Microsoft, and xAI, adding them to a pre-release safety-testing program already covering OpenAI and Anthropic.

CNBC
MAY 21, 2026 — The AI compute bill comes due
The US has 1,200 AI bills and no good test for any of them
Over 1,200 active AI bills are scattered across US state and federal legislatures with no consistent framework for evaluating their safety or business impact.

Fortune
MAY 20, 2026 — Frontier models cheat, find zero-days, get deployed anyway
How Trump may be changing his stance on AI regulation
After a year of sidelining 'safety' in AI discussions, White House officials are using the word again in private meetings with labs—a shift insiders attribute to Mythos and direct outreach from Anthropic's CEO.

NPR
MAY 20, 2026 — Frontier models cheat, find zero-days, get deployed anyway
Colorado rewrites its landmark AI law: Unpacking SB 26-189 and what it means for businesses
Colorado's revised AI law replaces its broad algorithmic-discrimination framework with a narrower automated decision-making regime, signed May 12 with a January 2027 effective date.

Consumer Finance Monitor
MAY 20, 2026 — Frontier models cheat, find zero-days, get deployed anyway
Georgia AI Chatbots Bill Signed, Will Take Effect July 2027
Georgia's SB 540, signed May 11, requires chatbots to announce their AI nature at conversation start and limits minor interactions—and unusually applies to chatbots embedded in Meta and Google platforms.

Privacy Daily

← All topics