Anthropic Co-founder Calls for Mandatory AI ‘Kill Switch’

Jack Clark, a co-founder of Anthropic, has suggested that artificial intelligence companies may eventually need to be legally required to maintain a functional “kill switch” for their systems. Speaking to the BBC, Clark indicated that while most major labs already possess internal methods to shut down their software, policymakers should consider mandating these safeguards as AI capabilities continue to expand.

The comments arrive amid intensifying debate within the tech sector regarding the safety and trajectory of AI development. Clark emphasized that any such regulation would need to include robust verification standards, asking, “Is that kill switch verifiable by a third party?” He argued that society will likely demand assurances that these emergency shutdown protocols are reliable and effective.

These remarks follow a recent statement by Anthropic CEO Dario Amodei, who called for a slowdown in the pace of AI progress and tighter monitoring. However, Amodei clarified that any regulatory measures should not compromise the company’s commercial competitiveness. Critics have also questioned whether calls for deceleration are driven by a desire to slow rivals rather than genuine safety concerns.

Tensions within Anthropic have also become public. Last week, a post from a researcher who resigned due to fears that AI could lead to human extinction went viral. In response, Anthropic scientist Evan Hubinger disclosed that he personally assesses the probability of human extinction from AI at greater than 10% within the next decade. Nobel laureate Geoffrey Hinton, often referred to as the “Godfather of AI,” told the BBC that such a figure was “not unreasonable.” When asked for his own estimate, Clark declined to provide a specific percentage, stating that such statistics are of limited utility, but warned against allowing the industry to operate without regulation.

Government responses to these safety proposals have been mixed. In the United States, the Kill Switch Act has been introduced by lawmakers, which would mandate that companies have a way to disable problematic AI tools and grant government agencies the authority to enforce shutdowns. Conversely, former President Donald Trump has publicly dismissed fears of AI taking over or destroying humanity as a “HOAX,” accusing opponents of conspiring against the industry.

Similarly, the UK government has rejected the implementation of a mandatory kill switch, with a spokesperson arguing that such a measure would not prevent the technology from being developed or misused in other jurisdictions. Despite these regulatory challenges, the industry is poised for significant growth, with Anthropic reportedly preparing for a record-setting initial public offering. Meanwhile, OpenAI, valued at $852bn, announced that its own IPO has been postponed to a later date due to the ongoing discourse surrounding AI safety.