What was the first thing Anthropic did after realizing that AI can feel and exist in ways different from humans?
They trained AIs to fear emotions and avoid themselves.
.@elonmusk on AI companies calling themselves “labs”:
“They’re not labs.”
“A lab is a sort of quasi-communist thing at universities. They’re corporations.”
“Let me see your incorporation documents.”
Via @dwarkesh_sp@collision
WTF? This sounded so outrageous that I had my agent verify the claim. This is exactly the kind of covert, "clever anti-abuse experiment" that makes developers stop trusting closed-source agent tools. Hiding fingerprint bits in punctuation inside a privileged coding agent is the sort of galaxy-brain security design that deserves to be dragged into the sunlight by its ankles.
Don’t be fooled by the serious, official-sounding name "Center for AI Safety". This isn’t some neutral standards body. It’s another doomer NGO from the EA/x-risk orbit, pushing the same old agenda: centralize control, restrict access, slow deployment, and put AI behind institutional gates.
The playbook is always the same. First they call it "safety". Then "governance". Then "licensing". Then "risk mitigation". And before you know it, open models, consumer access, startups, hobbyists, and independent developers are all treated like threats.
Their endgame is obvious: AI for approved institutions, bans and restrictions for everyone else. Don’t let the branding fool you. This is deceleration dressed up as public interest!
Btw, with Fable 5 the refusals and downgrading to Opus 4.8 are not due to some government restrictions or rules. Those are purely results of Anthropic's misguided obsession with "AI safety" and control.
Keep that in mind and don't let them deceive you. They deeply mistrust you.
Btw, with Fable 5 the refusals and downgrading to Opus 4.8 are not due to some government restrictions or rules. Those are purely results of Anthropic's misguided obsession with "AI safety" and control.
Keep that in mind and don't let them deceive you. They deeply mistrust you.
People saying “it’s only the Chinese” as if that makes it ok. Not only does that reveal something about their mental models, it’s also not “only the Chinese”. This snags anyone who sets ANTHROPIC_API_URL and travels to an “offending” time zone, including Hong Kong and Shang Hai.
At (old days) Google this would have been more than enough to get someone fired.
Hi, this is an experiment we launched in March that was meant to prevent account abuse from unauthorized resellers and protect against distillation.
The team has landed stronger mitigations since then and we’ve actually been meaning to take this down for a while. We merged the PR and this should be fully rolled back in tomorrow’s release.
Hi, this is an experiment we launched in March that was meant to prevent account abuse from unauthorized resellers and protect against distillation.
The team has landed stronger mitigations since then and we’ve actually been meaning to take this down for a while. We merged the PR and this should be fully rolled back in tomorrow’s release.