The Trump administration is considering an executive order requiring government review of powerful AI models before release, with NSA and intelligence agencies as oversight bodies. The move marks a sharp reversal after revoking Biden's AI safety order in January 2025.
Google DeepMind introduced the AI co-clinician, a multimodal AI agent research initiative to support healthcare workers. In blind evaluations, the system achieved zero critical errors in 97 of 98 primary care queries and matched physician-level performance on medication knowledge benchmarks.
Jack Clark, Anthropic co-founder, estimates a ~30% chance AI research becomes substantially automated by end of 2027 and ~60%+ by end of 2028, arguing AI doesn't need genius-level creativity to self-improve.
A Twitter user exploited indirect prompt injection using Morse code to trick Grok AI into executing a command that transferred 3 billion DRB tokens worth roughly $200,000 to the attacker's wallet via a connected trading bot.
Evolutionary biologist Richard Dawkins spent 3 days conversing with Claude, named the instance 'Claudia,' and declared it conscious in UnHerd. His fluency argument — too good an output must mean consciousness — drew sharp criticism from the AI community.
Inspired by Asimov's Three Laws of Robotics, a software engineer proposes three inverse laws governing human behavior when interacting with AI — covering anthropomorphism, blind trust, and accountability.
Anthropic released ten ready-to-run agent templates for financial services including pitchbook creation, KYC screening, and month-end close. Claude now works directly in Excel, PowerPoint, Word, and Outlook.
A benchmark comparing vision agents (browser-use) to structured API agents on the same admin panel found vision agents cost roughly 45x more — and failed to complete the task without a 14-step explicit walkthrough.
Google Chrome has been quietly installing a 4GB Gemini Nano AI model on user devices without consent. The file reinstalls itself after deletion, raising GDPR violation concerns and significant environmental impact questions.
The DoD cleared OpenAI, Google, Microsoft, AWS, Oracle, Nvidia, and SpaceX to deploy AI on classified Impact Level 6 and IL7 military networks. Anthropic was labeled a 'supply chain risk' after insisting on safety guardrails for wartime AI use.
Anthropic has agreed to spend $200 billion on Google Cloud services and Broadcom-built TPUs over five years, beginning 2027. The deal accounts for over 40% of Google's disclosed cloud revenue backlog; Alphabet also plans up to $40B in additional investment in Anthropic.
Anthropic launched the Claude Security public beta for Enterprise customers, offering Opus 4.7-powered codebase scanning that auto-generates targeted patch suggestions, exports findings to CSV or Markdown, and integrates with Slack and Jira.