Anthropic warns it may soon meet its own AI-speed-up threshold: keep a named human answerable

Anthropic said in an amended registration statement filed with the US Securities and Exchange Commission on 2 November that it "may imminently meet" the definition of "dramatically accelerated progress attributable to AI" in its own Responsible Scaling Policy. TechCrunch, which quoted the filing's risk-factor section at length on 2 November, reported that the filing says Anthropic's internal estimates and its expectations for future models, rather than its current models, point that way. The filing also states that Anthropic has not designated any of its current models as exceeding the threshold and continues to assess the underlying data.

The same section says Anthropic has in the past not identified, and may in future not identify, a significant acceleration in its rate of development until after it has occurred. It says the company does not anticipate being able to safely use Claude to fully autonomously design and build successive generations of frontier models without meaningful human supervision or understanding, and that it is considering updates to its Responsible Scaling Policy to keep humans in control of frontier model development. Anthropic first filed to list on Nasdaq on 26 October. The language is a risk warning to prospective investors about what may happen, not a statement that it has happened.

For a head teacher, a parent or a manager, the technical definition matters less than the plain point. A major AI developer is telling investors that its own rate of progress may be hard to see until after the fact. Tools that schools and workplaces adopt this term may change faster than their policies, training and assessment can follow, and the people who rely on them are not in the room when that happens.

Dan Fitzpatrick's reading is that this is a reason for steadiness, not alarm. Nothing in the filing says a current model has crossed the line, and curiosity should beat fear. But it is a disclosure that the pace can be hard to measure even for the people building the technology, which means the safeguard cannot sit with the supplier alone. It has to sit with the people using the tools: a named person who answers for how AI is used, and staff and pupils who can still do the thinking themselves. His principle is to outsource the doing, not the thinking, and the more capable the tools become, the more that distinction matters.

School leaders, educators, parents and business leaders can act now without waiting for any company to define its thresholds. Name one accountable person for AI use in your school or team. Review your AI policy on a fixed date every term rather than once a year, so it keeps pace with the tools. Ask each supplier in writing how it will tell you when a tool changes in a material way. Keep at least one task in every subject or role where people do the thinking unaided, and put AI literacy for staff and pupils ahead of any single tool.

Dan Fitzpatrick advises schools and education bodies on adopting AI and speaks and writes internationally on what it means for learning and work. His focus is practical: helping teachers and leaders decide what AI is for before the tools decide for them.

Dan Fitzpatrick is available for interview and comment on this story. He is available today for live or pre-recorded radio, TV, podcast and print interviews, with a broadcast-quality remote setup for live or pre-recorded down-the-line interviews. Contact: dan@theaieducator.io