Logan Graham
1.3K posts
Logan Graham
@logangraham
Head of the Frontier Red Team @anthropicai. 🌎 Make things radically good.
the present, moments ago Katılım Haziran 2009
8.1K Takip Edilen19.1K Takipçiler
@geoffreyirving Thanks for your exceptional work -- you have shaped the world.
English
A bittersweet announcement! For family reasons, I will be leaving AISI soon to move back to the Bay Area. I will be starting a new nonprofit alignment research org (more to come).
I will miss this place! Here are some reflections about my time at AISI. 🧵❤️
English
@TheStalwart Probably always relative, but I’m hopeful we can use models to increase the costs for attackers by quite a lot.
English
Cybersecurity is something I’ve literally never spent much time learning about. And probably I should change that. But I’m curious, is there such thing as provably secure software, or is security always relative to state of the art hacking technology?
Andrew Curran@AndrewCurran_
Mythos has cracked MacOS. It took five days.
English
"Within a year, Mythos will probably look quite dumb (relative to other new models)."
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: xbow.com/blog/mythos-of… UK AISI report: aisi.gov.uk/blog/how-fast-…
English
@sethbannon @AnthropicAI Likewise! Look out for more from us and partners soon. Obviously, we think this is a bigger discussion
English
@logangraham @AnthropicAI Appreciate the transparency! Important we talk about these things to prepare everyone for the new normal.
English
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: xbow.com/blog/mythos-of…
UK AISI report: aisi.gov.uk/blog/how-fast-…
AI Security Institute@AISecurityInst
Our cyber range results illustrate this step-up. Since our first Mythos evaluation, we received access to a newer Mythos Preview checkpoint. On a 32-step corporate network attack we estimate takes a human expert ~20 hours, this checkpoint completes the full attack in 6 /10 attempts.
English
It has been an honor helping here from the sidelines and even having seen some of this work first hand I don’t think I am ready for the implications and that scares me. Huge wake up calls are needed
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: xbow.com/blog/mythos-of… UK AISI report: aisi.gov.uk/blog/how-fast-…
English
@fluxxrider @AnthropicAI UK AISI previously tested a partially trained version. The latest results are on the actual Mythos Preview -- the model as it was on the day we launched Glasswing (on April 7th).
English
@logangraham @AnthropicAI is there a newer finetune of mythos? Why are there two versions of it being tested by AISI (additionally, METR used the “early” version)
English
@hendrycks Yeah. That's probably the scenario we think about the most. The entire question is then how to smooth the transition as much as possible.
(which I think could require some unprecedented innovations in security)
English
@logangraham Eventually could be a long time, especially in critical infrastructure, much of which is running on Windows 7.
English
I'm optimistic this eventually favors defense over offense. We wanted to start this transition cautiously.
I've honestly been inspired by what orgs have been able to do with Mythos. More to come!
Alex Albert@alexalbert__
With the help of Claude Mythos Preview, the Firefox team fixed more security bugs in April than in the past 15 months combined.
English
The team I lead is part of The Anthropic Institute. The way I think about the Institute is "applied weird blue sky research".
So far we've got a good track record of tackling some of the most important ideas -- cyber, self-improvement, robots, national security.
Anthropic@AnthropicAI
We’re sharing the research agenda of The Anthropic Institute, or TAI. TAI will focus on four areas: 1) Economic diffusion 2) Threats and resilience 3) AI systems in the wild 4) AI-driven R&D Read the full agenda: anthropic.com/research/anthr…
English
Some main learnings so far:
1) Use models in security today to get a glimpse of the future.
2) Start finding and fixing things.
3) Figure out how to scale it when more powerful models arrive.
Claude Security is a great way to do that!
English
Something I've seen from Mythos / Glasswing is that one of the things companies/orgs/maintainers need most is a first model-driven boost in finding + fixing vulns.
That's a big reason we're scaling Claude Security today for more users: x.com/claudeai/statu…
Claude@claudeai
Claude Security is now in public beta for Claude Enterprise customers. Claude scans your codebase for vulnerabilities, validates each finding to cut false positives, and suggests patches you can review and approve.
English
Also, if you're a security researcher / leader really motivated by the mission of "solve the whole AI cyber problem", you should apply to Anthropic.
We're looking e.g. for vulnerability researchers, senior security researchers and engineers, AI security research leaders, etc.
Logan Graham@logangraham
Privileged to help lead this. Thankful to our partners. Mythos is an extraordinary model. But it is not about the model. It's about what the world needs to do to prepare for a future of models that are extremely good at cybersecurity. This is the start.
English
@FFmpeg @aakashxsh Our policy right now is we try to send a patch with every disclosure (unless it can't be reasonably done).
We want to practice what we think is a gold standard to be most helpful. At least until we all figure out a better equilibrium.
English
Talk is cheap, send patches.
Alex Albert@alexalbert__
We released Claude Opus 4.6 just two months ago. Today we're sharing some info on our new model, Claude Mythos Preview.
English
@sporadica I mean, we think there are good arguments for it and we seriously hope so.
The transition period could be bad, though, if we don't work hard.
Also, no one truly knows.
English
@creatine_cycle @theo I totally missed you'd covered our work. Thought this was a good piece and well-calibrated thinking!
English
"my guess is that we are three to maybe nine months away from every piece of software we rely on being exploitable by most models, even open weight ones" - @theo
"we're about to hit a point where every single hospital, library and local city government and their google accounts are exploitable "by AI" and there is going to be terrible legislation proposed as a result"
Anthropic@AnthropicAI
Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software. It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans. anthropic.com/glasswing
English
I spoke to Anthropic execs about the new model, which they called a "reckoning" for cybersecurity.
They claim it has already found vulnerabilities in every major operating system and web browser, including some that "literally decades of security researchers" didn't find.
English
NEWS: Anthropic's new model, Claude Mythos, is so powerful that it is not releasing it to the public.
Instead, it is starting a 40-company coalition, Project Glasswing, to allow cybersecurity defenders a head start in locking down critical software.
nytimes.com/2026/04/07/tec…
English
Seeing this on Slack that day was one of the first "oh, I guess we're just seeing it now" moments for those who think about AI security
Trenton Bricken@TrentonBricken
Mythos sandbox escape and many more wild instances are in the Model Card
English
This release is also sort of a responsible disclosure.
Models are going to get better, and alongside that will come cheap, fast exploitation capabilities.
We need to prepare for that world.
red.anthropic.com/2026/mythos-pr…
English
Our team has been pointing Mythos Preview at every security task they can. It's really good. One big change is models of this capability class can write exploits -- sometimes sophisticated ones.
Mostly, we want you to know this may soon be the new reality.
English
Privileged to help lead this. Thankful to our partners.
Mythos is an extraordinary model. But it is not about the model. It's about what the world needs to do to prepare for a future of models that are extremely good at cybersecurity.
This is the start.
Anthropic@AnthropicAI
Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software. It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans. anthropic.com/glasswing
English