“There is not yet a viable scientific plan to solve risks from recursively self-improving AI. Please look up!”
Wang supports the warning that researchers lack a viable scientific plan for risks from recursively self-improving AI. She nevertheless believes she can reduce risks from inside a lab and explicitly respects Coxon and others who choose to leave.
- Speaker
- Anna Wang
- Capacity
- Anthropic employee; former Google DeepMind employee; explicitly personal capacity.
- Observed
- September 10, 2026
- Published here
- September 10, 2026
- Targets
- Anthropic, OpenAI
- Classification rationale
- Conditional defense of insider safety work, alongside agreement that the risk remains unresolved.
Surrounding material
Full context
Before the quotation
Wang identifies herself as a former Google DeepMind employee now at Anthropic and says the concern is common among her peers. She states that she is writing in a personal capacity.
After the quotation
In the next post, Wang says working inside the lab may be better for reducing risk, while strongly respecting and endorsing others including Jacob who think outside work is better.
Evidence
Complete source record
Full primary public post and available surrounding thread. Related context: https://x.com/a_nnawang/status/2097720576047710674 https://x.com/a_nnawang/status/2097720578254028909 Speaker and position scope are explained in the linked event.