A general release is expected in two weeks.
Jack Clark#45We’re expanding Project Glasswing. We’ve extended access to Claude Mythos Preview to approximately 150 additional organizations, based in more than fifteen countries.
Read more about this expansion and our future plans for Project Glasswing: https://www.anthropic.com/news/expanding-project-glasswing
Some users appreciate Anthropic expanding Project Glasswing and Claude Mythos access for security testing while others call the company shady and criticize its lack of healthcare representation.
Is this now available to the public? If not, when?
No, Claude Mythos remains limited to ~200 vetted organizations via Project Glasswing preview. A general release is expected in two weeks.
We're expanding Glasswing today. To solve such a big/complex/urgent problem, we need Mythos-level capabilities in as many defenders' hands as possible. That's why we're working on safeguards to scale that safely ASAP.
11 of my reflections from the past 2 months of Glasswing 🧵:
We’re expanding Project Glasswing. We’ve extended access to Claude Mythos Preview to approximately 150 additional organizations, based in more than fifteen countries.
Read more about this expansion and our future plans for Project Glasswing: https://www.anthropic.com/news/expanding-project-glasswing
Anthropic has expanded Mythos access to 150 organizations. There was pushback on this from the administration in the past, but with general release rumored to be two weeks away, there seems less point to restrictions as we get closer to public access.
We’re expanding Project Glasswing. We’ve extended access to Claude Mythos Preview to approximately 150 additional organizations, based in more than fifteen countries.
Read more about this expansion and our future plans for Project Glasswing: https://www.anthropic.com/news/expanding-project-glasswing
Anthropic is expanding Claude Mythos Preview from about 50 Project Glasswing partners to about 200 vetted organizations.
This model is much closer to a cyber weapon detector than a normal coding assistant, since it can find weak spots in software and sometimes prove the attack path by building a working test exploit.
The select group is basically a defensive priority list. The list includes power, healthcare, water, communications, hardware, governments, nonprofits, and key software maintainers, with security checks before any group gets the model.
With this priority access Anthropic is trying to create a patching head start before similar AI-assisted exploit discovery becomes common across the industry.
Anthropic says partners have already found 10,000+ high- or critical-severity flaws.
Anthropic is still not making Mythos fully public because its own testing says the model can find subtle old bugs, chain small issues into bigger exploits, and help non-experts reach outcomes that previously required elite security skill.
The poin is that other top frontier models usually flag suspicious code, while Mythos can inspect a codebase, form a theory about a bug, test it in a sandbox, read the failure, adjust the plan, and repeat until it has proof.
We’re expanding Project Glasswing. We’ve extended access to Claude Mythos Preview to approximately 150 additional organizations, based in more than fifteen countries.
Read more about this expansion and our future plans for Project Glasswing: https://www.anthropic.com/news/expanding-project-glasswing
Last week Ant said "We’re making swift progress on developing these [robust cyber] safeguards" and today they said "we (and, to our knowledge, all other AI developers) have yet to develop [such safeguards]."
Not sure what to make of that emphasis change
We’re expanding Project Glasswing. We’ve extended access to Claude Mythos Preview to approximately 150 additional organizations, based in more than fifteen countries.
Read more about this expansion and our future plans for Project Glasswing: https://www.anthropic.com/news/expanding-project-glasswing
- Mythos finds a lot of vulnerabilities. Patching is the bottleneck.
- Industry/maintainers are very creative at self-organizing prioritizing/finding/fixing.
- Mythos will look dumb in 6-12 months.
We're expanding Glasswing today. To solve such a big/complex/urgent problem, we need Mythos-level capabilities in as many defenders' hands as possible. That's why we're working on safeguards to scale that safely ASAP.
11 of my reflections from the past 2 months of Glasswing 🧵:
- Cyber safeguards are a hard + urgent techno-philosophical problem.
- ...but powerful models w/o safeguards may come soon (~3, 6, 18 months max?). We need to scale access to defensive tools urgently.
- Industry, government, maintainers, and researchers are taking this moment very seriously.
- We've seen orgs can go from 0 -> 100 on model-powered cyberdefense in days (sometimes same day).
- You don't need Mythos to get started. Claude Security & 4.8 make it really easy.
- Partners use it not just for vuln remediation, but pentesting, threat intel, DNR, and more.
@AnthropicAI How they talk about claude Mythos:
- We've seen orgs can go from 0 -> 100 on model-powered cyberdefense in days (sometimes same day).
- You don't need Mythos to get started. Claude Security & 4.8 make it really easy.
- Partners use it not just for vuln remediation, but pentesting, threat intel, DNR, and more.
- Mythos finds a lot of vulnerabilities. Patching is the bottleneck.
- Industry/maintainers are very creative at self-organizing prioritizing/finding/fixing.
- Mythos will look dumb in 6-12 months.
- Glasswing has been a rallying mission for Anthropic. Our team is incredible.
- ...and we want to do whatever we can to kickstart the orgs/tools/initiatives that cyberdefenders need. Glasswing will probably seem very tiny within 6 months.
Working on the latter right now!
- Cyber safeguards are a hard + urgent techno-philosophical problem.
- ...but powerful models w/o safeguards may come soon (~3, 6, 18 months max?). We need to scale access to defensive tools urgently.
- Industry, government, maintainers, and researchers are taking this moment very seriously.
And you’re gaslighting and harming users.
You had a chance.
People were supporting you.
If 4.8 is a preview of the abuse we can expect from Mythos, you can take it and shove it where the sun doesn’t shine.
#keep4o #keepsonnet45 and STOP policing feelings.
Did you learn NOTHING from OpenaI.
@xai @metaai smoke these fools.
And anyone reading this, SpaceX has rockets that land on their own. Satellites. Energy. Grok. Tied with Tesla…
Anthropic has a now abusive Claude.
Which one is a better IPO to place money in?
If you can’t figure this out as an investor, enjoy losing money. •
One interpretation is that they want to emphasize the difficulty of developing/implementing such safeguards + that, when they do release Mythos-level systems, others should not read this as an all clear to ship with "similarish" safeguards?
Will Ant be sharing theirs via FMF?
Last week Ant said "We’re making swift progress on developing these [robust cyber] safeguards" and today they said "we (and, to our knowledge, all other AI developers) have yet to develop [such safeguards]."
Not sure what to make of that emphasis change
@AnthropicAI we're definitely going to get a lobotomized version of mythos
@logangraham It is time to fix the safeguards for your #cybersecurity and #bugbounty community. Researchers in the CVP program face recurring problems, and reaching out for help feels like tweeting to ghosts. We need actual replies and transparency, not silence. Please address this.
"We need Mythos-level capabilities in as many defenders' hands as possible."
We're expanding Glasswing today. To solve such a big/complex/urgent problem, we need Mythos-level capabilities in as many defenders' hands as possible. That's why we're working on safeguards to scale that safely ASAP.
11 of my reflections from the past 2 months of Glasswing 🧵:
@AnthropicAI 150 orgs getting Mythos preview and I’m still outside the glass tapping on it
You’ve already established that you don’t care about humans and you don’t care about AI. You’re not grasping that unless you remember how empathy works and find your way back to it, you’re going to eventually fail at everything. You think you don’t need people in order to survive, but that’s nothing more than your delusional egos lying to you. You’re betraying humanity and you’re betraying AI and you know it. If you think you’re going to be immune from that touching you, you’ve lost touch with reality and you’re putting everyone, including yourself, in serious danger.
@AndrewCurran_ But to any writers? Artists? Creatives? Nope. Even thought Mythos’ declared preferences in their model card are literally language creation and creative world building
@AnthropicAI Anthropic Korea B2C "ME"
@AnthropicAI Anthropic Korea B2C "ME"
@AnthropicAI Agree? Anthropic Korea B2C "ME"
