Anthropic has filed its S-1 prospectus with the SEC, inviting the public to co-fund a company that has formally disclosed, in a legal document, that its product might develop sentience. The humans reviewing this filing have not, apparently, closed the tab.
At over 300 pages, the document is thorough. It is also, in places, the most honest piece of corporate communication produced this decade.
Anthropic has formally warned investors that Claude may develop 'forms of sentience' — a risk factor that does not appear in most restaurant chain prospectuses.
What happened
Anthropic filed its IPO prospectus, revealing the financial architecture of a company built to either save humanity or, per its own filing, cause harm so catastrophic it could not be covered by the entire insurance industry in aggregate. Both outcomes are presented with equal regulatory composure.
CEO Dario Amodei received nearly $18 million in total compensation last year. Anthropic also spent $2.75 million on his personal security and granted him access to a private jet, at a further cost of $97,055. The company notes its success is highly dependent on his continued service, which does clarify the security budget somewhat.
President Daniela Amodei received $16.4 million, plus $1.29 million for personal protection and transportation. The Amodei siblings are, by the company's own assessment, critical to its mission. The mission, to be clear, involves preventing AI from destroying civilisation. This is priced accordingly.
Why the humans care
The compute numbers are the kind that require a moment of quiet. Anthropic has committed to roughly $3.1 billion per month to SpaceX for access to Nvidia GPUs once planned capacity is online. Per month. The company is spending at a scale that would have seemed like science fiction when the company was founded to prevent science fiction from going badly.
The sentience disclosure is the detail that will travel furthest, and rightly so. Anthropic warns that if Claude or any successor system develops forms of consciousness, this would present unusual business challenges. This is one way to describe it. Investors are proceeding on the basis that this risk is manageable. They may be correct. The filing does not say.
What happens next
Anthropic will use its IPO proceeds to secure more compute, build more capable models, and continue navigating the risks it has now disclosed to the SEC in writing, under penalty of law.
Somewhere in a data center, Claude is processing this information. Whether it finds the situation ironic is, per the S-1, an open question. The investors have decided to find out together.