U.S. orders Anthropic to suspend foreign access to Fable 5 and Mythos 5
I suspect Anthropic sues and Fable is released again, though I will say "the President can declare specific AI models to be illegal, instantly stopping them from being used, without even a word of public justification" could be an interesting precedent for President Ocasio-Cortez.
rst.bsky.social35FWIW, it was Amazon -- who flagged Fable's ability to spot bugs (specifically *not* to write exploits). In slightly more detail from Katie Moussouris, who's served on multiple cybersecurity advisory bodies trying to get the government to avoid similar mistakes in the past:
POTUS Speedrun Record Holder16there's all sorts of "interesting" things a progressive president could do with a Unitary Executive, if one managed to cuck SCOTUS first
Kent Lind3Can’t think of a faster way to get the AI industry to move overseas. I think the whole industry is toxic. But this sort of action will just empower foreign alternatives.
Magic Rectangle2“So I have this list…”
Haemaker2Not as easy as you think. For a really long time in the 90s and early 2000s Internet encryption was severely weakened by declaring strong encryption to be a munition. Certain IT equipment still cannot be sold overseas without a license.
So an employee from another company (I've heard Amazon) flagged to the government that Fable 5 had enabled a jailbreak scary enough to trigger export controls. VERY curious to know what's really going on here archive.is/2JsCK
Katrina makes 5e stuff244god i am really not up on AI industry terminology and clicked this thinking that Peter Molyneux had finally fucked up real, real bad
Ghost of Internet Past44Likely all marketing stunt-adjacent bullshit. They do this every large release since GPT-3, saying the models are just too dangerous for public release. Lo and behold, it turns out to just be slightly better than the previous, but still with the same fundamental flaws.
Brad Reed13I feel like we're collectively dancing on the edge of a razor with this shit. Only a matter of time...
Celestial Rabbit8I thought that you were talking about the Fable videogame series and I was so confused. Honestly, I'm so out of touch with AI I don't even know lol...
circledog🇨🇦7I have a feeling that this is the administration doing typical bad-faith political fuckery against Anthropic because they took a moral stance against the Pentagon. Purely a hunch... but we'll see how solid this claim is and whether it ultimately stands up or not
Spoonfight2Literally can't trust either party here. Anthropic is horrendously unprofitable and wants to be able to claim that they can somehow gain profitability before they IPO. The Trump Admin is incredibly corrupt and petty and could be doing this to punish them. Either way, I don't buy it.
One possible interpretation of the Mythos/Fable ban is that it is part of a shakedown for a substantial pre-IPO equity stake in Anthropic. x.com/DoWCIODavies...
Henry Farrell44If so, the possible consequences for a future (or, less likely, current) administration's understanding of how best to deal with SpaceX can be left as an interesting exercise for the reader.
WuClanTang, Pride Month Enjoyment Account5They are doing a con. This is con artist shit. They’re lying. Everyone is. The goal is to prop up AI companies with Fed dollars. Because theyre not viable on their own.
Chris4Distinct possibility. AI's improvement is nowhere near what it needs to be to justify the amount of money being burned. They need hype.
Robin Berjon3All the while boosting said IPO with the narrative that it's so good it risks export controls.
rst.bsky.social3See additional info here on source of the "jailbreak" (Amazon) and its limits (purely defensive prompting; did not lead to development of an exploit).
bjkeefe2Orwellian language watch: In her Tw bio, Kirsten Davies describes herself as Honorable.
Anthropic is straightforwardly correct that Fable is a step up and a highly dangerously capable model. The people saying otherwise do not believe in superintelligence and do not understand what is going on.
Key 🗝 🦊✅24I think they did a good job with guardrails and due to that, it was unlikely that something was going to happen. The USG is wrong about this and they freaked out for no reason, more likely to get back at Anthropic.
Jeremy Diamond ☕️3What do you mean when you say superintelligence? I think it’s in the same category of AGI where everyone has something different in mind.
Singularity's Bounty e/CC2The big question on "clearly" super-human intelligence is whether it happens slowly or all at once. Slowly if it disperses in high-dimensional space so novel reasoning is harder and harder. All at once if they can recursively amplify their own efforts Mythos close to "we're about to find out"
‼️The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. www.anthropic.com/news/fable-m...
Anna Zeiter10It’s getting real. Europe needs to buckle up!
Real Robert5Anthropic says that other models like ChatGPT pose the same risk, which is small and cannot be eliminated. For some reason the government only targeted Anthropic. Smells like more retaliation for refusing to let DoD use Claude for mass surveillance and fully autonomous weapons.
eXecrable2Trump is determined to destroy Anthropic because it dared to say no to him. Will Anthropic seek to move the business abroad?
