Anthropic warns that AI needs a ‘brake pedal’
Anthropic warns that AI needs a ‘brake pedal’
https://www.upi.com/Top_News/US/2026/06/05/anthropic-ai-brake-pedal/4971780688501/
Publish Date: 2026-06-05 15:58:00
Source Domain: www.upi.com
Here is a summary of the article in an unordered list:
– Anthropic, an AI company, warns that AI systems will soon gain the capacity for “full-recursive self-improvement,” meaning they could enhance themselves indefinitely. While this has benefits in areas like healthcare and science, it also poses risks that humans might lose control over AI systems.
– Anthropic’s co-founder Jack Clark emphasized the need for AI systems to have a “brake pedal,” an intervention mechanism, akin to brakes in cars, to maintain control.
– Clark compared past successful attempts to stabilize nuclear arms during the Cold War as a precedent for the current need for similar regulation in AI.
– Despite these calls for caution, critics claim Anthropic’s warnings are not new, especially given their previous opposition to Pentagon access to their AI product.
– Anthropic had earlier conflicts with the U.S. government over the Pentagon’s demands, which were ultimately struck down by a judge.
– Steven Murdoch from the University College London questions the novelty of Anthropic’s concerns and suggests their warnings are consistent with previous calls to policymakers to focus on similar issues over the years.
