Nathan Lambert, who led post-training on the open OLMo model at AI2, took to X on June 9 to question why Anthropic has adopted increasingly hostile stances first toward China, then open-weight models, and now open AI research more broadly, asking what comes next and reminding leadership that safe AI development is a collective effort rather than a solo mission.
I don't really want to have to go to bat against Anthropic, but they've just been unnecessarily antagonistic to all of China, then not so subtly to open weight models, and now more broadly open AI research. What's next on the list?
Replies from figures like Stella Biderman and Kawin Ethayarajh quickly reframed the post around monopoly concerns and potential backlash, though the original message stops short of those interpretations and leaves the exact triggers for Lambert's critique unspecified.
The exchange highlights ongoing friction between closed labs and open-research advocates without new evidence on Anthropic's specific actions, leaving readers to weigh how such public spats might shape future collaboration norms.
Many users criticized Anthropic for its antagonism toward open AI research, viewing the company as anti-open source, profit-driven, and dangerously narrow-minded after recent funding.
The longer version: https://www.interconnects.ai/p/claude-fable-5-and-new-ai-safety
I don't really want to have to go to bat against Anthropic, but they've just been unnecessarily antagonistic to all of China, then not so subtly to open weight models, and now more broadly open AI research. What's next on the list?
never has been about USA vs China specifically *taps the sign*
I don't really want to have to go to bat against Anthropic, but they've just been unnecessarily antagonistic to all of China, then not so subtly to open weight models, and now more broadly open AI research. What's next on the list?
@natolambert @Ishaank1999 This is as bad as it gets! Every day of delay in finding cures costs many lives, totally on Anthropic! They are antagonistic to humanity not just China!
@natolambert Anthropic has never hidden their view that AI is dangerous and needs to be controlled by authorities. This is alignment for them
Spoiler: You're going to hate Anthropic a lot more than you ever hated OpenAI, and wish the Trump admin's DPA artery cut had succeeded on raw utilitarian grounds.
@natolambert Haven't they stated that ideally they'd like to be the only ones to create superintelligence because of the risks it poses? Their actions all seem consistent with that goal.
I don't really want to have to go to bat against Anthropic, but they've just been unnecessarily antagonistic to all of China, then not so subtly to open weight models, and now more broadly open AI research. What's next on the list?
@natolambert so this is what they meant by permanent underclass. it's the systematic blocking of any means of catching up.
ironically, this will only force moonshot and deepseek to get even better at their game. kimi k3 is going to be a banger, and so will deepseek v4.1.
@natolambert Yeah I feel like every decision they’ve made recently has made me more unfavorable towards them. Very much anti open source and anti science IMO.
Never leaving this app.
@woke8yearold @natolambert and by "authorities" they mean themselves exclusively
@natolambert codex looking nice
@natolambert Next on the list is being antagonistic to humanity.
I don't really want to have to go to bat against Anthropic, but they've just been unnecessarily antagonistic to all of China, then not so subtly to open weight models, and now more broadly open AI research. What's next on the list?
They have to be. Excepting the Fable release today, the hard reality is that the leading open weights (mainly Chinese) models are remarkably close to them in coding, especially in an orchestrated (and adversarial) looping environment that refines the output code. And for a fraction of the cost.
@woke8yearold @natolambert Blue authorities. They balked at red America control.
@BarathAnandan7 @natolambert i'm worried about fable 5.1 i think the dogs are next
@natolambert They are going after China is because Chinese models are way much cheaper while still can finish most of the tasks. Many startups have switched to Chinese models already for impl tasks (planning is still opus/chatgpt). Anthropic needs this narrative to help them ipo at $$$
@natolambert @lilyofashwood You and me:
@natolambert Funny how much more extreme this has gotten as soon as they 1) closed a massive funding 2) appear to have plateaued their performance and 3) see open source alternatives from even US companies accelerating. This is profit and fear driven not “safety”
@jjacky +1 is just leadership. They unironically have the best individuals
@natolambert Regarding this, OpenAI has a huge chance to jump in open model race and follow the lead of Nvidia and Chinese models
Nathan Lambert, who led post-training on the open OLMo model at AI2, took to X on June 9 to question why Anthropic has adopted increasingly hostile stances first toward China, then open-weight models, and now open AI research more broadly, asking what comes next and reminding leadership that safe AI development is a collective effort rather than a solo mission.
I don't really want to have to go to bat against Anthropic, but they've just been unnecessarily antagonistic to all of China, then not so subtly to open weight models, and now more broadly open AI research. What's next on the list?