Cesar Fernandez

Head of US State & Local Government Relations @AnthropicAI

[Cesar Fernandez @CesarFernand3z]

@zebulgar @rabois Welcome 😎

[Cesar Fernandez @CesarFernand3z]

@davidrgoldberg Yes for sure!

[Cesar Fernandez @CesarFernand3z]

😎

[Cesar Fernandez @CesarFernand3z]

Thank you, @GovPritzker!

[Governor JB Pritzker @GovPritzker]

Illinois is leading the nation in holding Big Tech accountable.

As AI systems impact people’s lives, we need safeguards in place.

I look forward to signing SB 315 and working with the legislature so that AI, when used, is used responsibly.

[Cesar Fernandez @CesarFernand3z]

As AI models grow more powerful, this kind of enforceable accountability matters more than ever.

Thank you @DanielDidech_IL, @SenEdlyAllen, @GovPritzker, @ILAttyGeneral and others for your leadership. Illinois lawmakers have set a new standard, and we hope others build on it.

[Cesar Fernandez @CesarFernand3z]

With bipartisan support, Illinois is on track to become the first state to require independent, 3rd party audits of large frontier AI developers' safety practices. I was proud to testify for @AnthropicAI in support of these important safety requirements.
https://t.co/57yNR3CrFh

[Cesar Fernandez @CesarFernand3z]

@ericwu01 This is very cool. Congrats on the launch.

[Cesar Fernandez @CesarFernand3z]

@hpierrejacques we have air conditioning here

[Cesar Fernandez @CesarFernand3z]

YES!

[americanair @AmericanAir]

Your scrolling just got an upgrade. Coming in 2027, @Starlink to American flights. ✈️

A more seamless, high-speed in-flight connection — built for streaming, browsing, gaming, and staying connected in real time, gate to gate.

Want to learn more? Head to our Newsroom. https://t.co/1ilXrMDRjD

[Anthropic @AnthropicAI]

Patching these vulnerabilities will make us safer. But the software industry will need to adapt to the volume of vulnerabilities that models like Claude Mythos Preview will be able to find.

We discuss this in our initial update on Project Glasswing: https://t.co/3cSgHHZXgG

[Cesar Fernandez @CesarFernand3z]

@karpathy Welcome, @karpathy!

[Cesar Fernandez @CesarFernand3z]

@M___Morgenstern @TJVillamil @claudeai Yes agreed

[Cesar Fernandez @CesarFernand3z]

@nottombrown This tweet is 10/10

[Cesar Fernandez @CesarFernand3z]

Small businesses drive 44% of US GDP and anchor every local economy. @AnthropicAI we are launching Claude for Small Business, free AI training, and a 10-city tour from Chicago to Baton Rouge to Baltimore to bring AI to the entrepreneurs who power our communities: https://t.co/bLu5lCClbf

[Cesar Fernandez @CesarFernand3z]

@travisk 👀

[Cesar Fernandez @CesarFernand3z]

Excellent interview of our CFO, Khrisna Rao, by @patrick_oshag. If you listen to 1 podcast this week, make it this one👇

[Logan Graham @logangraham]

A lot of people have been wondering about Mythos, Glasswing, and the vulns we / our partners are fixing. Today, I’m excited for us to start sharing more. (For context, I lead Glasswing @AnthropicAI.)

Two independent evaluations this week—from XBOW and the UK AISI—confirm what we've been seeing internally: Claude Mythos Preview is a step change in autonomous cybersecurity capabilities. We need to start preparing fast for a world of models with this level of capabilities.

The UK AI Security Institute tested the model we shipped at the launch of Project Glasswing and found Mythos Preview is the first model to solve both of their end-to-end cyber ranges, including one (Cooling Tower) which no model had ever cleared. But attackers (and defenders) have sophistication & cost constraints – Mythos is also the only model that clears every one of their tasks estimated over 8 hours under their deliberately low 2.5M-token cap.

XBOW tested it on their offensive security benchmarks, finding "token-for-token, unprecedented precision." It's the only model to succeed at subtle V8 sandbox work.

Other Glasswing partners shared similar stories. In a few weeks of testing, Mythos Preview has helped them find many thousands of (estimated) high + critical severity vulnerabilities, sometimes double what they'd normally find in a year.

I don't share this to boost Mythos. In fact, this is not about Mythos. It’s about preparing for the coming world of models being better, faster, cheaper, and more creative than some of the best human experts at dual use capabilities. Clearly, we need them supporting defenders as widely as can be done safely – and especially the least resourced ones.

Within a year, Mythos will probably look quite dumb (relative to other new models). And others may release openly available or unguardrailed models of Mythos-level capabilities.

We started Project Glasswing because capabilities like Mythos Preview's won't stay rare, or stay in careful hands. We are bringing it to defenders as fast as we responsibly can, while working to figure out, for example, the right safeguards and patching & disclosure processes.

Also, to be clear, compute has never been a limiter in our rollout.

Expect a fuller update on our Glasswing work in the coming days.

XBOW report: https://t.co/Mumtbf3kE3

UK AISI report: https://t.co/vBgqz0AeKJ

[Cesar Fernandez @CesarFernand3z]

We now have great insight into what Claude is thinking when you ask it a question or prompt it to do something for you. This is remarkable 👇

[Anthropic @AnthropicAI]

New Anthropic research: Natural Language Autoencoders.

Models like Claude talk in words but think in numbers. The numbers—called activations—encode Claude’s thoughts, but not in a language we can read.

Here, we train Claude to translate its activations into human-readable text.

[Sarah Heck @SarahKHeck]

We’re excited for continued engagement with community members and elected officials in Memphis. What we hear will shape our approach as a neighbor. Thanks for your partnership @mayorpaulyoung.