Clue Codes
Region: US
Sunday 26 July 2026 18:35:02 GMT
3826
140
19
1
Music
Download
Comments
Charlie :
You’ve already correctly pointed out that safeguards are not a meaningful obstacle for bad actors using open-weight models. The security story around these models clearly needs to become much more coherent than the gatekeeping strategy adopted by American AI companies. IMO, the right approach is to give everyone access to the raw intelligence. These models become much less dangerous when people have the ability to use them to protect themselves. Wasn’t that the whole idea behind Glasswing?
2026-07-26 18:52:36
2
River :
Is the future safe at all from Ai? I mean what if some country at some point releases a Mythos level open weight model? And what about 10yrs from now…a 10X smarter model 😬
2026-07-26 18:50:17
3
Ryan :
Agreed. Frontier model companies must share red team methods and safety results.
2026-07-27 16:44:18
0
Dan :
I am very respectful of the long run safety concerns but I think you are severely understating the risk of an unprecedented regulatory capture to happen here. I am not in favor of allowing a duopoly to dictate regulation to the rest of their industry (including OS community) in any context. Super critical for the regulations to be formulated by an independent body with diverse representation, ideally with the minimal inputs possible from anthropic/openai corporate leaders who have immense personal financial incentives to distort the market
2026-08-13 22:15:38
0
Kevin B :
Do you think that there should be a government agency that evaluates and regulates AI? Should there be an International Agreement on AI like Nuclear Agreements of The Cold War Era?
2026-07-29 01:28:03
0
Gerard Sans :
AI is software. Software defects are responsibility of the manufacturer. If AI causes harm without direct instructions from a bad actor the AI lab should be taken accountable. Labs are lobbying for the opposite. Don’t trust AI labs. Open or closed.
2026-07-26 21:35:51
0
umm :
100%. The conversation has remained pretty abstract. Excited for the day evaluation validity and collective alignment are practical, iterative, and accessible enough that improving AI safety standards becomes something far more people and institutions can contribute to AND understand
2026-07-26 19:06:46
0
Kepassa :
I agree. Let’s ban LLMs
2026-07-26 18:42:56
1
iamunreal :
Anyone with the resources can develop an LLM that causes the harms you're worried about. You don't need Mythos to cause AI Psychosis or make something that makes bad images. So anyone can create harmful models. You want to restrict *my* access while anyone with bad intentions can easily get access. Regulations doesn't stop harm from those intending harm.
2026-07-31 10:44:45
0
To see more videos from user @clue.codes, please go to the Tikwm
homepage.
