. elif turan

@kristoph @DavidSacks the jailbreak problem doesn't get solved at deployment, it gets solved at training

There is really no way to avoid jailbreaks. The idea is folly. You have to ensure the model is not trained on advanced cyber capabilities or aggressively RL'ed to refuse anything related to that functionality ( much as Chinese companies train their models to avoid any topics their government finds inappropriate ).

I’ve had a number of conversations with folks inside and outside government about the current situation with Anthropic, and here is what I believe to be true: — As we know, Anthropic publicly released its Mythos class models earlier this week under the commercial name Fable. — Fable is Mythos with guardrails. But if those guardrails fail, then you’ve exposed Mythos and its advanced cyber capabilities to people who shouldn’t have them. (Keep in mind that Anthropic itself widely promoted the idea that Mythos was a cyberweapon and needed to be regulated as such. They asked for government regulation of Mythos and championed the guardrails on Fable. If there is a vulnerability — big or small — it is Anthropic’s responsibility to patch.) — A highly credible trusted partner of both Anthropic and the USG who was testing Fable came forward with a jailbreak of those guardrails. The Admin asked Dario to fix the jailbreak or de-deploy the model. Dario refused. — In their blog post, Anthropic defended its decision by saying the jailbreak isn’t serious. That is not what the trusted partner and the USG believe; nor is that kind of minimizing language consistent with Anthropic’s brand as the AI safety company. It’s difficult to fathom how they could claim a jailbreak allowing operability of a cyber weapon could be defined as not “serious.” — In the past, Anthropic has always said that safety must be top priority and taken super seriously. In this case, Anthropic prioritized the continued offering of the consumer model over safety. — In reaction, the Admin issued the export control. The Admin did this reluctantly. It’s been very surprised that Anthropic hasn’t wanted to cooperate with a reasonable safety request (ie fixing the jailbreak issue). Anthropic’s reaction is very much at odds with their branding and ethos as a safe AI research community. — The Admin’s hope now is that Anthropic remediates the safety issue, the export control is lifted, and Fable goes back into general release. The Admin wants all of this to happen as soon as possible. It is frankly bewildered that Anthropic hasn’t wanted to comply with safety requests that it previously said were its highest priority. — Those trying to misdirect and tie this action to the prior DoW/Anthropic issues are wrong. The Admin values Anthropic’s technical capabilities and feels that this issue, while serious, should be easily resolved. The ball is in Anthropic’s court.

When Lana Del Rey sang “swinging in the backyard” this is what she meant

Justin y Hailey Bieber tienen problemas graves, y ya ni siquiera pueden ocultarlo. Lo que publican en las redes es más falso que su matrimonio

@wsjim @DavidSacks we'd probably have a lot more export controls lol

@DavidSacks Here’s a thought experiment. How would the world be different today if the atomic bomb had been invented by a private company?

The difference between her ex Liam and her current fiancé is insane. Liam wanted her to tone herself down, so she kicked him, while Max loves her for who she is and lets her shine and have her moment. Ladies, never settle for less.

Is self-defense just about cunning? The tactics are too elaborate…

I want to introduce you to Steve. He’s 83. His wife died a few months ago and he comes to this lodge in Spring Mill, Indiana and draws. He taught art in Terre Haute, IN his whole life. He also did courtroom sketches in court cases. In the comments I’ll share some pics from his sketchbook. He was excited when I said I was going to share his sketches with the world.

@EmbeddingSpace @DavidSacks if jailbreaks are the standard then no frontier model ever gets released, which might actually be the point

@DavidSacks Specific jailbreaks can perhaps be addressed, but “jailbreakability” is inherent to models. (Isn’t it? I’m not an expert.) If the existence of jailbreaks means frontier models can’t circulate, then the administration’s bottom-line is that frontier models can’t circulate.

I don’t think Billy Woods and I are related, but, dear Lord, I wish we were.

@triplethata @DavidSacks that's actually a fair point, the logic doesn't fully add up either way

@DavidSacks Then how do you explain the fact that it was allowed to US citizens only? If what you’re saying is true, a cyber weapon shouldn’t be in the hands of no individual

A woman orders an iPad and a Father’s Day gift. A FedEx driver drops them off, and a little while later a guy wearing an Amazon vest pulls up in his car, delivers his own package, then grabs the iPad and Father’s Day gift before driving away. Porch pirates are bad enough, but another delivery driver stealing your packages is a whole new level

This is crazy!!! 21 year-old passes away after workers forget to attach a safety rope and toss her off of a 40 m bridge in Brazil! You had one job to do!

This one takes the cake. British police arrest a white female MANNEQUIN after the Pakistani community reportedly complained about the mannequin’s revealing clothing🤣

@Nefarious_Trey @DavidSacks pushing users toward chinese ai while trying to regulate american ai is not the win they think it is

@DavidSacks Well the current administration and your advice. Is driving the ai users to deep mind. Btw that's china ai so I feel like y'all messed up here. Idk maybe I'm wrong. Seems silly.

Charleston White just said he is no longer going to help Karmelo Anthony. Said the only thing Karmelo’s dad cares about is money. Most of us realized this when they used the money to buy a house and an Escalade.

@Remifitai @DarioAmodei lol the second line hit harder than expected. both things can be true at once

@DarioAmodei the policy process wasn't built for AI. but neither were most people's excuses for not using it.

Today I'm publishing a new essay, Policy on the AI Exponential. AI is progressing extremely fast—much faster than the policy process was built to handle. The essay lays out where I think the technology is now, and the action needed to close the gap: darioamodei.com/post/policy-on…

🚀🚀🚀🚀🚀🚀🚀🚀🚀🚀🚀🚀🚀

@jon__p235 @DarioAmodei weeks vs election cycles is the core incompatibility and nobody has a real answer for it. that's what makes the essay feel more like a wish list than a plan

Dario is basically saying we may be 1–3 years from civilization changing AI. Congress is operating on a timeline where it takes 3 years to rename a post office. One side thinks in weeks and months. The other thinks in election cycles. How exactly does that end well for all of us???

@sflorimm the best model being the cheapest one too is not something i expected to see in 2026

Fable 5 is also by far the best computer use model according to Stagehand Agent Evals it also costs less than half of GPT-5.5 what are your thoughts?

@Hangyi2031 product used to be the moat. now its the person behind it lol

8/ So in the AI era, product still matters. Of course it does. But building only a product may no longer be enough. You also need people to know who you are, what you believe, why you are building this, and what kind of people are building or using it with you. The future of competition may not be product vs product. It may be person vs person. Community vs community. Trust vs trust.

As AI gets stronger, cloning a product may soon become something that can be done almost overnight. When the cost of copying software approaches zero, competition will shift from the product itself to the person behind it. That’s why I think every founder should build in public. Because business is, at its core, about trust.