Anthropic co-founder Jack Clark is calling for lawmakers to consider making AI 'kill switches' mandatory, allowing for the shutdown of AI systems that pose risks. In an interview with the BBC, Clark emphasized the need for verifiable kill switches that can be checked by third-party entities.

Anthropic scientist Evan Hubinger has estimated the chance of human extinction from AI at over 10%, although Clark does not share this figure, stating that such statistics are not particularly useful. Nevertheless, Clark warns against leaving the AI industry largely unregulated, stating that 'we are rolling dice with immense risks.' He believes it's not too late to act, as the industry has not yet reached the most dangerous stage.

Clark's proposals include bringing in outside auditors to AI companies, establishing common safety standards backed by industry and governments, and facilitating dialogue between major powers like the US and China. Anthropic has already agreed to allow third-party evaluators into its lab within weeks, with Clark naming METR as one of the testing groups. This move is expected to cause some friction within the company.

President Donald Trump has rejected calls to slow down AI development, calling AI safety fears a 'HOAX.' Clark argues that countries risk losing their lead in a technology when an accident destroys public trust, citing nuclear power as an example. China has also criticized the slowdown push, but Clark notes that Chinese models are currently six months to a year behind Western models and will face the same risks within months.

Clark leads a team of economists at Anthropic, which has developed a new economic model. This model predicts that a significant jump in AI capabilities will bring a big rise in GDP, but also unprecedented white-collar unemployment. Clark warns that this is the default path if governments do not act to regulate the industry.

The US has proposed a bill called the Kill Switch Act, which would require companies to maintain a means of shutting down problematic AI tools. Some government agencies would also gain powers to order these tools shut down or restricted. While the UK government has rejected the idea, other industry leaders, including OpenAI's Sam Altman, Elon Musk, and Google DeepMind's Demis Hassabis, have expressed support for parts of Anthropic's call.

This article was written with the assistance of AI.
News Factory APP - agentic news to boost your SEO & AEO.