Not everyone in AI is buying doomsday predictions

The BBC has found that not all employees at major artificial intelligence companies share fears that the technology poses an existential threat to humanity. Several current and former workers at firms including OpenAI, Meta and DeepMind said they reacted with amusement, rather than alarm, to a recent wave of high-profile warnings.

The discussion was sparked by claims last week from former Anthropic employee Jacob Coxon, who suggested a group of AI agents based on models that do not yet exist could create and deploy a biological weapon, though he did not explain how this would happen in practice. His warnings, which called for a slowdown in AI development, were echoed online by other industry figures, including some at Anthropic, OpenAI, DeepMind and Elon Musk's xAI.

A former OpenAI employee, now at another AI company, told the BBC their first reaction was surprise, saying such claims tend to be vague or rely on large leaps in reasoning. Rishub Jain, who founded AI safety research firm Sampura Research this summer after seven years at DeepMind, said the mood among industry workers had been "a little jokey," noting the debate over existential AI risk is not new. Meta data scientist Colin Fraser wrote on social media that there is no real evidence AI models would inevitably pursue goals leading to human extinction.

Despite the light-hearted tone, workers said they take seriously more immediate risks, such as users or hackers bypassing AI safety guardrails, and growing ethical concerns over AI's expanding role in military applications. Several noted that no independent safety researchers appear to be embedded within AI labs yet. Anthropic announced on Friday it would bring in evaluators from Faculty, an AI firm owned by Accenture, which is also a business partner of Anthropic, though no timeline was given. A separate recent incident involving OpenAI and Hugging Face has also been widely described as a wake-up call for the industry over security risks.