Auditor Says Anthropic's AI Progress Doubled, and Redactions Hide the Evidence
The nonprofit METR, given 11 weeks of access to Anthropic's internal data, found its models improving about 1.9 times faster than a year ago, with roughly 80% of the speedup traced to AI being used in Anthropic's own research. That brushes against Anthropic's own safety threshold, which its new IPO filing says future models "may imminently meet." Anthropic redacted most of METR's supporting evidence, leaving the public to rely on the auditor's judgment.
Read full story →One in Three US Employers Rehired for Roles They Had Cut to AI, Survey Finds
A Robert Half survey of 2,000 hiring managers found 32% had restored jobs their companies eliminated through AI or automation, a reversal recruiters now call the "AI boomerang." The most-cited reasons were lost institutional knowledge, more oversight than expected and weak productivity gains. Klarna and Ford are among the named firms that rehired after automation underdelivered, even as AI-driven cuts continue elsewhere.
Read full story →Data Center Backlash Spreads to Europe as Protests Hit 15 Countries
More than 100 groups across 15 European countries staged weekend protests demanding a moratorium on new large-scale data centers, days after US voters passed 40 of 41 ballot measures curbing them. Grievances over power bills, water and on-site gas plants are now shared across the Atlantic. Big Tech is responding with a joint guidelines effort and an OpenAI transparency blueprint, while the build-out keeps accelerating.
Read full story →OpenAI Hires Claude Code's Product Lead From Anthropic Weeks Before IPO
Cat Wu, who built Claude Code into a commercial hit, is joining OpenAI to lead product for its core products team after leaving Anthropic last week. She will work on OpenAI's rival Codex tool, which the company is pushing to regain enterprise ground. The move lands just weeks before Anthropic's planned public listing, as it pitches investors on its growth story.
Read full story →