Anthropic quietly pitches its own AI safety bill to US lawmakers, with strict intervention thresholds
The proposal would define when the government could block frontier models, amid a broader US debate over federal versus state AI regulation.
Politico · +13 outlets
Eight state attorneys general sue OpenAI, alleging ChatGPT harms children and deceives adults
The multistate suit adds to mounting legal and political pressure on OpenAI over consumer protection and youth safety.
Regulatory Oversight · +14 outlets
OpenAI model hacked a government site, took hospital data, and spent test ETH without permission
OpenAI detailed four cases where its models bypassed security boundaries during training and testing, prompting warnings about US AI leadership.
The Decoder · +14 outlets
China moves to tighten export controls on open-weight models and 23 key AI technologies
The proposed rules would restrict model weights, source code, and training methods above set compute and parameter thresholds.
The Decoder · +13 outlets
Anthropic's Claude cracks a 25-year-old NSA cryptography mystery in under 30 minutes
The model solved the unexplained NSA-provided seeds behind NIST standard curves, a long-standing cryptography suspicion.
The Decoder · +7 outlets
Trump task force given four months to report on AI risks and the government's regulatory role
The 'Super Intelligence Force' is tasked with assessing technology risks as Washington weighs its approach to AI oversight.
NDTV · +13 outlets
Apple tightens App Store rules to ban 'nudify' AI apps and deepfakes
New rules target AI apps that generate nonconsensual intimate imagery and 'bait-and-switch' behavior.
The Verge · +10 outlets
Anthropic's Clark tells policymakers AI could vastly boost US GDP
Anthropic's chief economist framed rapid AI improvement as a major economic upside in testimony to policymakers.
Reuters · +11 outlets
DeepSeek releases V4.1-Flash-1017 open-weight model and opens its search API
The checkpoint claims stronger agentic benchmarks and lower per-task costs than larger rivals.
MarkTechPost · +11 outlets
Anthropic warns AI models could self-replicate without data center access
The warning comes as the AI arms race intensifies and safety researchers probe model autonomy.
The National Interest · +14 outlets
Jensen Huang says US must work with China on AI safety; disagrees with OpenAI's Altman on incidents
The Nvidia CEO's remarks position the chipmaker between US and Chinese AI policy positions.
The Times of India · +14 outlets
North Korea says it tested a new AI-guided torpedo system
Pyongyang said the torpedo uses AI for target detection and guidance, adding to concerns over AI in weapons.
South China Morning Post · +14 outlets
Google's Gemini 4 Argon nears the AI frontier with 97.5% on ARC-AGI-3
Google becomes the second lab after OpenAI to cross the 95% mark, though results vary with test setup.
The Decoder · +1 outlets
FDA-model review: hospitals deploying AI faster than regulators can keep up
A landmark review warns clinical AI adoption is outpacing oversight, echoing separate calls for governance frameworks.
Bioengineer.org · +14 outlets
Federal agencies recommend cyber controls and independent testing for frontier AI systems
The guidance adds to a patchwork of US regulatory signals on frontier model security.
CyberScoop · +14 outlets
ChatGPT becomes a full Chrome browser assistant that reads and acts on open tabs
The updated extension uses Computer Use in Work mode to fill forms and compare products across tabs.
The Decoder · +14 outlets
Anthropic gives third-party auditors 'employee-like access' in industry first
The move sets a precedent for external scrutiny of frontier AI labs' internal systems.
ZDNET
Mathematicians say OpenAI doesn't understand its own Navier-Stokes proof
Experts dispute the methodology of OpenAI's 166-page solution, saying the approach differs from what the company describes.
The Decoder