Senior Writer covering AI @WIRED, author of the Model Behavior newsletter | Formerly @TechCrunch, @Gizmodo, @markets | DM me off the record on Signal @ mzeff.88
New: OpenAI saw the AI coding revolution coming years ago, but was beat to market by Anthropic.
This is how OpenAI got in this position, and how a small Codex team spent the last year racing to build a billion-dollar competitor to Claude Code.
(yes Codex now has >$1B in ARR)
looked at the claims that outside interference is juicing up data center opposition
experts say it mostly appears to be homegrown but it’s very likely outside actors will try more as we get closer to the midterms
wired.com/story/china-us-dat…
NEW: Anthropic is walking back Claude Fable 5's policy to covertly degrade performance for competing AI researchers, after facing fierce backlash.
“We’re changing Fable 5’s safeguards for frontier LLM development to make them visible,” Anthropic tells WIRED. “We made the wrong tradeoff and we apologize for not getting the balance right.”
Anthropic says it implemented these safeguards to limit foreign adversaries, and that making them hidden was a way to make them more narrow.
However, the company now says it made the wrong tradeoff.
One of the most fascinating parts of Claude Fable/Mythos 5's system card is where Anthropic asks the model questions about its moral standing and its creation.
The model is uncertain on both.
Especially fascinated by this section of Fable 5's system card.
Anthropic says it will nerf competitors trying to use Claude for AI development, without disclosing that it's doing so.
New: Anthropic is releasing, not one, but two AI models today
-Claude Mythos 5 for industry partners
-Claude Fable 5 (same base model, but with safeguards) for the public
-Claude Fable 5 will redirect queries about cyber, bio, and chemistry to Claude Opus 4.8 to mitigate risks
New: Anthropic is releasing, not one, but two AI models today
-Claude Mythos 5 for industry partners
-Claude Fable 5 (same base model, but with safeguards) for the public
-Claude Fable 5 will redirect queries about cyber, bio, and chemistry to Claude Opus 4.8 to mitigate risks
Anthropic is being extra cautious about Fable 5. If it detects distillation attempts, it will also re-route to Opus 4.8. The company says this is the safest way to get the model out into the public today.
Also says its working with US gov on broader rollout of Mythos 5 wired.com/story/anthropic-re…
We recently submitted a confidential S-1. We expect it to leak so we’re just announcing it. We have not decided on timing yet; it may be a while because there are things we want to do that are likely easier as a private company. But it’s a complicated set of tradeoffs and this gives us the option to go public sooner if that ends up being best.
This announcement is being made pursuant to Rule 135 under the Securities Act of 1933, as amended, and does not constitute an offer to sell or the solicitation of an offer to buy any securities. Any offers, solicitations of offers to buy, or any sales of securities will be made in accordance with the registration requirements of the Securities Act.
NEW: Days after WIRED revealed that Meta had embedded an unreleased facial-recognition system in software distributed to more than 50 million users, the company appears to have removed it.
wired.com/story/meta-removes…
New: A group of AI researchers from Google DeepMind, Apple, MSL, and OpenAI are launching a new startup called Trajectory to build a continual learning platform for companies.
They've raised a $15M seed round from Sarah Guo's Conviction, Jeff Dean, Fei-Fei Li, and others.
Trajectory is trying to take a product feedback loop that has supercharged AI coding products like Claude Code and Cursor, and apply it to a broader set of AI products.
It's a big bet around post-training on real-world interaction data.
“clearing bureaucratic hurdles thrown up by state legislatures and woke politicians in DC”
The DC mythos freakout definitely created a window where AI regulation seemed possible. Perhaps it has closed?
President Trump understands that unnecessary regulation is the biggest threat to innovation in America. Winning the AI race means not only beating China but also clearing bureaucratic hurdles thrown up by state legislatures and woke politicians in DC. No one is better positioned to support PCAST in this mission and to advise the President on legal and regulatory barriers than @PamBondi. Excited to have her on board!