AI doomer warnings put Anthropic’s business interests under scrutiny
TechCrunch’s Equity podcast examines claims of existential AI risk, the incentives behind them and the more immediate harms facing workers, the environment and the climate.
Warnings that artificial intelligence could pose an existential threat to humanity are prompting fresh debate over what the claims mean for AI companies, investors and regulators.
TechCrunch’s Equity podcast examined the latest controversy after AI researcher Jacob Coxon reportedly resigned from Anthropic, saying leading AI companies were “gambling with our lives”. An Anthropic alignment lead later wrote: “We really do earnestly believe AI could kill all humans!” The lead added that their personal estimate of the chance was “>10% within the next decade”.
The podcast treated the claims with caution. The more-than-10% figure was presented as a personal view, with no methodology provided, while the alignment lead’s use of “we” was questioned. The discussion also referenced “P(doom)”, a term used for estimates of catastrophic AI risk.
Participants debated whether warnings about increasingly capable models and agents behaving unexpectedly reflect genuine concern, capability signalling or marketing. They also discussed whether such statements could align with business interests as Anthropic considers a potential initial public offering.
That speculation extended to how existential-risk language might appear in Anthropic’s prospective S-1 filing and whether it could affect valuation. No filing, timing or valuation impact was confirmed in the discussion.
The episode also turned to regulation and nearer-term risks, including potential effects on labour, the environment and the climate. TechCrunch said the recording took place before Anthropic chief executive Dario Amodei published a plan for more cautious AI development.
