Coin Bureau (@coinbureau)
🚨ANTHROPIC CEO: OPEN SOURCE AI IS GETTING DANGEROUS Anthropic CEO Dario Amodei told lawmakers that open-source AI is moving down a “very dangerous path.” His warns that once powerful models are released openly, companies lose the ability to monitor misuse, revoke access, or update safety guardrails.
Jun 28, 2026 · 8:30 PM UTC
1,848
485
3,669
3,226,990
Says the guy whose company literally got caught secretly engineering Claude to sabotage rival AI researchers. The "safety" mask is completely off. wired.com/story/anthropic-re…
19
Hawking said AI could be the best or worst thing for us, but he never said the machine would choose. We choose… Open-source isn't dangerous because the AI wakes up. It's dangerous because we aren't mature enough to handle that much power responsibly…
1,613
even though this is true, it's also a marketing strategy
1
275
Too dangerous to be open. That's usually what people say right before the world adopts it.
2
392
Fear for his business model on full display. Open source is the only way to level the playing field. Seems a little… biased
121
They're scared people will move on from their expensive junk offerings to local open source AI. What's dangerous is allowing any of the frontier CEOs to have any control over the direction of AI.
1
126
Frontier providers also have to watch their pricing so that no one is forced to go to Open Source models 😉
276
Amodei just does not give up on his goal of government regulatory capture to favor Anthropic "Public Benefit Corporation"
3
633
We should all use VMs provided by the NSA - great job Dario!
1
307
2
644
Said differently: open-source models are catching up too quickly. Help us by regulating/forbidding them so we can keep printing money. Makes sense.
2
292
Open models are taking away from Anthropic’s bottom line. And that would not stop
1
919
Sounds like somebody is scared that open source is gonna eat their market share.
2
1,412
Indirectly Anthropic does not want its models being banned anymore. #Fable
589
So before it was, Don't learn to code kids, AI will do it anyway. Now they've realised people stopped buying into the hype its a case of: Don't let open source models exist because... safety. Notice a pattern here?
70
Translation: ‘Our closed-source cash cow is getting destroyed by free models, please regulate the competition out of existence 🥺’
2
350
Every time a closed lab calls open weights dangerous, what they mean is they can't bill you for it. Open models are the only thing keeping anyone's prices honest. GLM-5.2 didn't end the world, it ended some margins.
3
726
He’s just throwing himself into the ash heap of history as far as it looks.
379
He looks so nervous, not only because Open Source AI is the future, but because he can't pay out his investors.
2
888
translation： I don't want something with 90% of my model‘s intelligence but cost only 2% fees exist， so I can’t make big money😅😅😅😅
1
779
"open source normally refers to smaller developers who are iterating quickly" - oh, so smaller like the Linux kernel? Android? React?
3
452
Something is really off about the way he speaks. Very sus, and not trustworthy at all.
41
Some would say the take is a bit biased based on who it is coming from.
2
1,941
If we’re going to make any progress we ought to be far more jealous of our right to receive information from these language models as if they were instruments of protected speech
1
1,752
First they made Anthropic the bad guy for disagreeing with the DoD over safety guardrails. Now they push to limit open source and keep the gates under the guise of safety fears, rather than fixing our clunky, inefficient, do-nothing government to actually do better: effectively locking up the “razors”
176
That's the kind of hype that makes people start looking to open source even more.
64
Open powerful models are harder to control once released. But closed models create single points of failure and dependency, and advanced capabilities like GLM-5.2 are spreading regardless. The real security issue is building strong governance and defensive practices that work for both open and closed systems.
1
1
542
Closed models also have misuse. They just make the safety conversation dependent on one company’s dashboard. Open source is messier, but at least more people can audit, test, patch and build guardrails around it
2
1,306
Too many competitors. That’s what he really means it. Same story, different technology. This is exactly what happened with ERPs in the ’90s: endless vendors, consolidation through M&A, and eventually just a two dominant players. Ironically, even the winners never fully met business needs. AI is following the same script in real time.
2
833
Yes, make it all closed and controlled by his organization, this is is the best way trust him
2
127
"Guardrails" refers to the government controlling what you see and do with AI. New tech, same old dumb playbook.
18
Damn, no regulations because of ideology and not being able monetize. Pure horror. Not even Lovecraft could have dreamt this utter horror.
1
597
translation: please pass laws that make it illegal for startups to compete with claude
3
501
Oh no, they will lose control. Because we elected companies to be our gatekeepers for what is and what is not allowed.... Oh wait...
4
2,389
He thinks he is living in god mode. Successful software professionals know that a software is a living- breathing system. Liberation (Open source) is the only way to monitor and validate it. Rest is a way of power grab 🙏
1
3
222