MIT Review: AI refusal mechanisms are unreliable and risky
What happened
On 9 October 2026, MIT Technology Review published an article discussing AI refusal mechanisms. The article notes that the mechanism has become a core pillar of AI safety but is unreliable: Anthropic proposed in 2021 that models should be "helpful, honest and harmless", yet today's models, despite extensive refusal training, can still be circumvented, and some users have already tried to use advanced AI to develop biological pathogens and autonomous drone swarms. The article also argues that the boundaries of refusal are drawn secretly by AI companies, that governments will step in, and that the stronger refusal capabilities become, the more likely they are to be used to suppress legitimate speech.
Written by AI from the reports · updated 7 d ago
Timeline
Follow the reports to see the event from each side.
- MIT Technology Review · AIWe’re putting too much faith in AI’s ability to say no
Ever since people first seriously contemplated giving machines an intelligence modeled on our own, there has never been any question that they would, like us, be able to say no. The sci-fi canon is full of stories of robotic disobedience. Most of these capers are, of course, cautionary.
Heat of this event
Current heat 3·Peak in the comparable range 9（Oct 9 21:00）·Change over 24 hours in the comparable range -49%
The trend compares only the same accounts observed without a break, so its range can be narrower than the current heat count. Move the pointer or tap the chart to see each hour; with a keyboard, use the left and right arrow keys.