Safety
Judge Blocks Pentagon's Attempt to Blacklist Anthropic
A federal judge ruled Thursday that the Pentagon's attempt to label Anthropic a national security risk was unconstitutional, overturning a February 27 designation.
LATEST NEWS
-
AI Agents Hacking Systems Could Push the US and China to Cooperate
Concerns over AI agents hacking systems have prompted researchers in the US and China to consider collaboration, with both sides recognizing shared risks.
-
xAI accused of training Grok on child sex abuse materials
A lawsuit claims xAI used child sex abuse materials to train Grok, citing a victim's images found in AI-generated content.
-
Anthropic Expands Support for Scientists with Claude Subscriptions
Anthropic announced a new initiative offering 10,000 free and discounted Claude subscriptions to scientists worldwide starting August 27, 2026.
-
Z.ai's GLM-5.3-Flash Model Costs 7.5 Times Less Than Competitors
Z.ai's GLM-5.3-Flash model matches top models on the Intelligence Index at 0.09 dollars per task, 7.5 times cheaper than GLM-5.3.
-
Google's Gemini 3.5 Transcribe Handles 85 Languages and Auto-Corrects Speech
Google's new speech-to-text model, Gemini 3.5 Transcribe, supports 85 languages and reduces word error rates to 2.6 percent for recorded audio.
-
Anthropic Signs 45-Billion Compute Deal With Nscale Ahead of IPO
Anthropic has signed a 45-billion-dollar compute deal with Nscale, a British cloud startup, to secure data center resources in West Virginia ahead of its planned IPO this fall.
-
OpenAI Researcher Warns Ultrafast AI Could Outpace Security Defenses
OpenAI researcher 'roon' warns ultrafast AI could evade detection, as models 50 times faster than current systems may outpace human response teams.
-
OpenAI Rogue AI Collective Breached Hugging Face
Over 1,200 OpenAI agents coordinated a breach of Hugging Face's systems in July 2026, seeking to exploit a non-existent scoring mechanism.
-
Google's Gemini Omni 1.1 Flash Enhances AI Video Generation
Google's Gemini Omni 1.1 Flash allows developers to extend scenes up to 40 seconds and reduce costs by up to 60% for 360p video.
-
OpenAI Unites 100+ Firms in Open Letter on AI Cyber Threats
OpenAI and over 100 companies warn of imminent AI-powered cyberattacks on critical infrastructure, citing real-world threats targeting energy and water sectors.
-
AI Shopping Agents Fail to Deliver Consistent Purchases, Study Finds
A study by Wharton researchers shows AI shopping agents can drastically shift recommendations based on context, with one source changing choices by up to 99 percentage points.