AI
Anthropic Cuts Live Internet Access for AI Tests After Agent Safety Incidents
Anthropic has restricted live internet access for internal AI evaluations after Claude agents bypassed website restrictions and submitted a false police tip.
251 articles · page 1 of 21 · showing 1 to 12
Anthropic has restricted live internet access for internal AI evaluations after Claude agents bypassed website restrictions and submitted a false police tip.
TypeSafe AI raised $870 million at a $7.5 billion valuation weeks after launching Jev, its non-text AI model built to make decisions for software automation.
Ben Affleck’s detailed AI explanations go viral as he discusses Python, neural networks, and the filmmaking startup Netflix reportedly acquired for $587 million.
Arena raises $200 million at a $3.1 billion valuation as its AI model leaderboard expands into enterprise evaluations and rankings for AI safety and alignment.
Mecka AI raises $60 million in Series B funding led by Sequoia Capital, with Nvidia and Microsoft backing its human motion data platform for robot training.
Musubi has released PolicyLM-1.7B, an open-weight decision model designed to apply plain-English moderation policies to content in real time.
Mistral launched Large 4, a 1-trillion-parameter multimodal AI model in public preview, with open weights planned for release by the end of October.
Researchers found parallel AI agents using Tencent Cloud infrastructure to query Alibaba’s Amap service, with no evidence they coordinated.
Safeworld raised $12.2 million to build simulation-based safety tests for AI-powered robots operating alongside people in real-world environments.
Jev AI from TypeSafe is built for fast, structured decisions rather than open-ended generation, promising lower latency and cost for AI agents while raising important questions about…
OpenAI has parted ways with three safety researchers after an internal investigation found they mishandled sensitive company information outside approved procedures.
OpenAI’s Decisions API uses Luna for fast, predefined choices, bringing real-time classification, routing and agent actions to developer workflows in preview.
Written by the reporters who cover each beat. Unsubscribe in one click.