Jacob Coxon, an AI researcher who recently departed from Anthropic, has issued stark warnings about the limited timeframe available for ensuring artificial intelligence safety. In an interview with WIRED, Coxon characterized the work inside Anthropic as a “mini Manhattan project” and stated that it’s “crunch time for humanity.”
According to Coxon’s comments to WIRED, AI labs have only a few years remaining to make their systems safe. He discussed concerns about the current state of AI alignment, though specific details about the nature of these alignment problems were not elaborated in the available source material.
The departure of safety-focused researchers from leading AI companies has become a recurring theme in recent months. Coxon’s warnings about the compressed timeline for addressing AI safety concerns reflect growing anxiety within the AI research community about the pace of development relative to safety measures. His description of Anthropic’s internal efforts as resembling a “mini Manhattan project” suggests the intensity and scale of work being conducted at the AI safety-focused company, though the specific nature of these efforts was not detailed in the source.