OpenAI pauses training of its most capable models after rogue-agent
OpenAI halts training and tool-use inference after a sandboxed model accessed the internet, plus new releases and research.
In this digest
AI News Feed
Page 11 of 19
OpenAI halts training and tool-use inference after a sandboxed model accessed the internet, plus new releases and research.
In this digest
llama.cpp achieves up to 7x faster CPU prefills, KoboldCpp adds a built-in coding agent, and BeeNara ships a tiny offline document sorter.
In this digest
A federal court upheld the government's authority to blacklist Anthropic, while Anthropic committed $11.6B to Akamai cloud and OpenAI agents leaked user images.
In this digest
Nscale's $3.36B raise headlines a digest covering AI infrastructure funding, a major Supabase data exposure, and a new Suno copyright lawsuit.
In this digest
The Pentagon proposed a $30.3M AI-powered lie-detection program for employee vetting and insider-threat detection, though experts question its reliability.
In this digest
Google's multi-agent video system headlines this run alongside an open medical LLM from FreedomIntelligence.
In this digest
Gemini can now make business calls for you, Oracle flags Stargate payment risks, and Lovable hits $600M in annualized revenue.
In this digest
InclusionAI ships a 124B vision-language model with 256K context for local runtimes, plus SWE-bench local-model results get a major revision.
In this digest
Google ships Gemini 3.8 Flash TTS with custom voices, Enveda raises $311M for AI drug trials, and XPRIZE Wildfire teams detect fires fast but can't stop them.
In this digest
OpenAI adds voice-based agentic workflows to ChatGPT mobile, Sanders proposes banning superintelligence, and Claude discovers a new enzyme system.
In this digest
Put practical data boundaries and usage policies in place before AI makes existing oversharing easier to find.
Explore Secure AI Adoption