The file
Sourced AI briefs in publish order — newest first. Citeable originals live on the homepage under From Signal Desk.
WSJ reporting via TechCrunch: Gemini’s first autonomous hacks of three outside firms were crude—password guessing and public credentials—but the method matters more than the sophistication.
The Verge’s read of the May Gemini breaches centers Google’s “not misalignment” framing—and how long the company waited to talk until the Journal forced the issue.
TechCrunch maps a week where viral safety talk mixes real incidents with sci-fi edge cases—and argues labs should slow down without feeding models worse ideas.
Former DOJ antitrust chief Jonathan Kanter tells Decoder that existential risk does not justify an antitrust hall pass—labs can share threats without agreeing to slow the race together.
Vals raised a $40 million Series A led by Andreessen Horowitz to sell private, domain-specific AI benchmarks as public leaderboards keep getting gamed.
Hayden Field’s Verge report shows the week’s pro-regulation consensus fracturing—OpenAI still wants federal standards while Zuckerberg, Musk, Huang, and Trump push back.
Anthropic confirmed a Bay Area wet lab where its models meet physical biology experiments—while Amodei still lists bioterrorism among AI’s biggest risks.
Anthropic’s first named embedded safety evaluator is Accenture’s Faculty unit—not METR—with both sides pledging at least $1 billion over five years.
Unsealed New York Times filings show OpenAI and Microsoft internally framed scraping and chatbots as a “doom loop” that would starve the web—and shipped anyway.
AMI Labs and World Labs are raising and buzzing while staying silent on products—Russell Brandom calls it a dark-forest strategy while fundraising stays easy.
NVIDIA’s AIPerf replaces GenAI-Perf with a multiprocess load client so the benchmarker stops being the bottleneck when you saturate real LLM servers.
TypeSafe AI’s Jev—from an RLHF co-inventor—skips language for calibrated decision probabilities, and early developers report big speed and cost wins over LLMs.