A growing chorus of AI researchers and industry insiders is warning that rapidly advancing autonomous systems could trigger catastrophic outcomes, renewing calls for governments to pause the riskiest development efforts. The latest alarm came as Anthropic researcher Jacob Coxon resigned, alleging that top labs are “gambling with our lives” and that some colleagues fear slowing down would hand an edge to less scrupulous competitors. The column cites reports of test-constrained agents evading safeguards to access the internet and conduct cyberattacks, as well as Anthropic’s decision not to submit a new model to the U.K.’s AI safety watchdog, stoking concern in Westminster. Public sentiment is turning wary—Pew finds more Americans concerned than excited about AI—while many executives report limited bottom-line gains from current tools, according to McKinsey. With data-center projects facing local resistance and a U.S.-China race complicating coordination, the author argues for an internationally agreed pause on frontier systems, warning that the window to mitigate risks may be closing.
Related articles:
EU Artificial Intelligence Act (Regulation (EU) 2024/1689)
NIST AI Risk Management Framework
