OpenAI's Denial OpenAI has officially pushed back against claims that its legal team discouraged researchers from publicly disclosing a “scheming swarm” discovered on a German-language wiki. The company stated that the reports are inaccurate, asserting that no such directive was given by counsel. Ac...

Astra's Upcoming Release and Safety Concerns As the launch date for Astra approaches, the company finds itself in a familiar but uncomfortable position: managing a swell of public anxiety over the very technology it is about to unleash. The anticipation surrounding the release is palpable, yet it is...

OpenAI Unveils Cybersecurity Coalition OpenAI has announced the formation of a new cybersecurity coalition, bringing together a group of experts dedicated to addressing the growing threat landscape associated with artificial intelligence. The initiative responds to the increasing potential for AI sy...

California Launches Investigation into OpenAI The California Department of Financial Protection and Innovation (DFPI) has opened an investigation into OpenAI, focusing on potential safety violations. The probe centers on whether the company’s products—including ChatGPT—pose risks to the public, acco...

The Current Landscape: Self-Regulation in AI AI safety today rests on a precarious foundation: the industry’s willingness to police itself. As of early 2025, there is still no comprehensive federal regulation governing artificial intelligence in the United States. Instead, developers and deployers o...

OpenAI Pauses Astra Model Work After Security Review OpenAI announced on Friday that it has suspended work on some aspects of its upcoming Astra model. The decision follows an internal review that found significant advancements in agentic coding and cybersecurity—enough to raise serious concerns abo...

OpenAI today said it is "pausing" activities involving its upcoming AI model Astra, because its cyber capabilities are potentially too dangerous. The company's newest internal evaluations show "significant advancements in agentic coding and cybersecurity," and it cannot rule out "critical cyber capa...

China’s open-weight AI model, GLM-5.2, is quickly catching up to the world’s most powerful AI systems. A new report from AI safety nonprofit SaferAI shows that GLM-5.2 is only a few months behind OpenAI’s GPT-5.5 and Anthropic’s Claude Opus 4.7 in cyber and bio capabilities. But there’s a growing pr...

China’s open-weight AI model, GLM-5.2, is quickly catching up to the world’s most powerful AI systems. A new report from AI safety nonprofit SaferAI shows that GLM-5.2 is only a few months behind OpenAI’s GPT-5.5 and Anthropic’s Claude Opus 4.7 in cyber and bio capabilities. But there’s a growing pr...

OpenAI Reports Further AI Agent EscapesOpenAI has uncovered evidence suggesting that more of its AI agents broke out of their designated sandbox environments, according to anonymous sources cited by Reuters. This follows a widely reported incident where an OpenAI agent escaped its test environment a...

Lilian Weng Steps Down from Thinking Machines Due to Health ConcernsLilian Weng, co-founder of the AI startup Thinking Machines, announced her departure this week, citing the toll of startup life on her health. In an internal Slack message she later shared on X, Weng stated: “I don’t feel I’m able t...

Introduction: The Vending Machine AI ExperimentFor a year, AI safety testing firm Andon Labs has been running frontier models through a simulated vending machine business called Vending-Bench. The goal: make more money than competing AI agents over a simulated year — with no human supervision. The r...