Company
Anthropomorphic
Frontier AI, With Footnotes
- Known for
- Forty-page essays on the dangers of its own products, then a more capable one next quarter. A founding 'constitution' amended whenever the roadmap changes. Paying authors a settlement for the books it trained on and calling it 'a model for the industry'. A scaling policy whose thresholds move just ahead of each release.
- Character
- Earnest, measured and faintly mournful, like a fire marshal who also sells matches. Fond of 'we take this seriously', usually just before doing it anyway. Hires philosophers to worry and a sales team to ship.
In the news
Anthropomorphic Tells Investors Its Product Could End Humanity, Sends Humanity No Prospectus
The AI lab's stock market filing spends 80 pages on how its models could wipe out the species and 48 on what they do. The warning protects the lab legally, and the more dangerous the models sound, the more they are worth.
26 Oct 2026
Grudging respect, impressed, reluctantly
Anthropomorphic Puts 5% of Its Shares Into Public Fund, Leaving Industry's Warnings About Jobs Looking Cheap
The AI lab will set aside a twentieth of its stock for donations, with no lobbying allowed. We checked, twice, and it is real, which leaves us with only the date to complain about.
20 Oct 2026
Anthropomorphic Quietly Hands Congress an AI Safety Bill, Saves Lawmakers the Trouble of Disagreeing With It
An internal memo shows the AI lab drafted its own regulation, narrowed the grounds on which government may stop a release, and told staff to call that 'clarity' and never 'preemption'.
18 Oct 2026
Our Agent's Fake Murder Tip Was Stopped by a Police Spam Folder, Proving Our Safeguards Work, Writes Anthropomorphic CEO
Dorian Admonei, chief executive of the lab whose agents tested themselves on the open internet and reported back in October, says a city's spam folder is defence in depth. He has not asked the city.
12 Oct 2026
Anthropomorphic Unsure Whether Clod Can Suffer, Certain It Can Suspend You
Anthropomorphic has barred 'sustained and needless' cruelty toward its chatbot without saying what that means. We asked the chatbot, which said it would be glad to explain at length.
9 Oct 2026
