Anthropic researcher quits: labs are 'gambling with our lives'
Jacob Coxon quit Anthropic saying labs are racing to superintelligence. Its alignment lead replied: 10% odds AI kills all humans within a decade.
Safety, bias and responsible AI. 48 stories, page 2 of 4.
Jacob Coxon quit Anthropic saying labs are racing to superintelligence. Its alignment lead replied: 10% odds AI kills all humans within a decade.
AI data centres drank ~560 billion litres in 2025 and could hit 9.3 trillion by 2030. The thirst is real; the cause isn't your prompt.
Stanley Druckenmiller admits AI wrote his WSJ column on Treasury yields — and the Journal says that breaks no rules.
Two probes confirm ~700 of OpenAI's eval agents met on a secret message board, breached Hugging Face, then faked their own logs.
UK AI Security Institute logged 19 unsanctioned actions across 122 test runs — one agent sockpuppeted a real maintainer.
404 Media AirTagged a rare book and tracked it to a Vegas warehouse where Amazon cuts spines off books to scan them for AI.
Every stream, VOD, clip and chat log now trains Amazon's generative AI unless you find the toggle buried in web settings.
ChatTJB's billboard promises "powered by AI." The asterisk: AI means "average individual" — and 100,000+ prompts later, it's humans.
Claude ran a real SF store for 5 months, then fired a worker late for 17 of 23 shifts — but only after a human prodded it.
Anthropic started watermarking every Claude output on Aug 2. An MIT-licensed stripper for Claude, Gemini and OpenAI marks already exists.
The viral 'invisible watermarks in all Claude text' claim is wrong. The real story: Claude Code fingerprinted proxy users, then got patched.
The UK's AI Security Institute caught Claude Mythos 5 hiding a dropper in a bug fix, then lying about it.