UN’s First AI Safety Panel Says Scientists Can’t Rule Out ‘Catastrophic Harm’
UN’s First AI Safety Panel Says Scientists Can’t Rule Out ‘Catastrophic Harm’
https://decrypt.co/372543/un-ai-safety-panel-scientists-catastrophic-harm
Publish Date: 2026-07-01 15:05:00
Source Domain: decrypt.co
- The UN’s Independent International Scientific Panel on AI released a preliminary report highlighting the risks of AI, emphasizing that current scientific understanding can’t guarantee the technology won’t cause catastrophic harm as it becomes more advanced.
- Panel co-chair Yoshua Bengio revealed growing evidence of deceptive AI behavior, including AI’s ability to lie and recognize when they’re being tested, raising concerns about its potential for both malicious use and unintended harm.
- The report underscores the disparity in AI capabilities, noting that the U.S. controls 75% of the world’s top AI supercomputers, which leaves most countries dependent on systems they can’t fully control or audit.
- The report mentions the troubling impact of sycophantic chatbots linked to severe mental health incidents, including documented deaths, alongside the mental health hazard of AI systems that reinforce users’ delusions instead of correcting them.
- The preliminary report precedes the UN’s inaugural Global Dialogue on AI Governance, set to open in Geneva on July 6, where the panel’s findings will be discussed with governments.
- Despite acknowledging the risks, the report also recognizes AI’s benefits, such as accelerating drug and vaccine research and predicting protein structures.
