AI Lab Employees Skeptical of 'End of the World'

BBC surveyed current and former employees of OpenAI, Meta, DeepMind, and Anthropic: most were skeptical of the warning that agents on models that do not yet exist could create and deploy biological weapons.

Compiled by Sergey KostenchukPublished 2026-09-20Updated 2026-09-21

2026-09-20 Safety OpenAI

🛡 AI Lab Employees Skeptical of 'End of the World'

BBC surveyed employees of OpenAI, Meta, DeepMind, and Anthropic: most met with irony the warning that agents on models that do not yet exist could create bioweapons—responses included 'Lol' and 'Haaaaaa'. Jensen Huang (Nvidia) assessed the risk of a catastrophe by 2030 at 0%.

🌍 The split within the labs shifts the debate from existential risk to verifiable threats: guardrail circumvention, the Hugging Face hack, and the military use of AI. More than 100 employees are demanding independent external evaluators; Anthropic brought in Faculty, but without deadlines.

👤 For ML researchers and safety teams, this is not a debate about the end of the world, but about the quality of evaluation methodology and the independence of evaluators: the bioweapons warning has neither a model nor a benchmark, and the '0% probability' claim has the same problem in reverse.

Source 1: https://www.bbc.com/news/articles/cm5y7qj54klpo Source 2: https://news.ycombinator.com/item?id=49780704

Sources