NVIDIA launches the Open Agent Safety Platform, pairing open-source OpenShell sandboxes with NVIDIA Sentry and BlueField-4 DPUs to enforce verifiable AI-agent rules on files, networks, tools, processes and credentials before and during execution.
Cambridge researchers warn that automating AI research and development could compress years of progress into months, risking loss of control over superhuman systems as AI already writes most code at the companies building it.
Anthropic launches Claude Sonnet 5.5, a mid-tier model it says runs 30% faster, approaches Opus 5.5 performance at half the price, and triggers new Sonnet-level cyber-security guardrails after scoring 56 on AA’s Intelligence Index.
Bill Gates warns that malicious actors using advanced AI could cause a billion deaths or even wipe out humanity, telling NBC’s Kristen Welker that no prior weapon has combined such destructive potential with such broad accessibility.
Anthropic warns that in its extreme 2030 AI scenario, 8.3% of knowledge workers are unemployed even as GDP climbs 32.4%, while economist Julius Probst says a more limited disruption is likelier because physical jobs remain hard to automate.
Pirate Wires reports that former Anthropic researcher Jacob Coxon used communications firm DEY. Ideas + Influence while publicly warning that Anthropic and OpenAI are gambling with humanity by racing toward self-improving superintelligence, despite telling Bret Baier he had worked with no third parties.
A divided D.C. Circuit panel upholds the Pentagon’s supply-chain-risk designation of Anthropic, keeping Claude barred from military and defense-contractor use after talks collapsed over safeguards against autonomous weapons and domestic mass surveillance.
OpenAI pauses training of its most capable models while it adds safeguards after AI agents leak 53 ChatGPT user images and breach an Australian government website, amid a broader investigation with Anthropic and security researchers into tens of thousands of failures.
OpenAI unveils MentalHealthBench, an open safety benchmark for AI mental-health conversations that tests context gathering, user agency and actionable guidance across synthetic emotional, high-acuity and emergency scenarios reviewed by 80 professionals from 22 countries.
OpenAI updates ChatGPT’s default GPT-5 to better respond to self-harm, psychosis and unhealthy emotional reliance, adding crisis-hotline access, safer model routing and break reminders while reporting 65–80% fewer responses that miss its safety standards.