Editorial illustration for Discord Users Access Anthropic's Mythos AI Tool Without Authorization
Hackers Breach Anthropic's Mythos AI Preview Tool
Discord Users Access Anthropic's Mythos AI Tool Without Authorization
Anthropic called its Mythos AI too dangerous to release. That was the official story. On Discord, a different narrative emerged.
A handful of users, sifting through a leaked dataset from the training startup Mercor, made a simple guess. They guessed where the model might live online. According to Bloomberg, they were right.
This wasn't a sophisticated hack. It was a deduction. Now, a tool designed to expose critical security flaws is itself exposed, its digital door barred by little more than a secret web address.
Discord Sleuths Gained Unauthorized Access to Anthropic's Mythos AI Tool Anthropic's Mythos Preview AI model has been touted as a dangerously capable tool for finding security vulnerabilities in software and networks, so powerful that its creator has carefully restricted its release. But one group of amateur sleuths on Discord found their own, relatively simple ways--no AI hacking required--to gain unauthorized access to a coveted digital prize: Mythos itself. Despite Anthropic's efforts to control who can use Mythos Preview, a group of Discord users gained access to the tool through some straightforward relatively detective work: They examined data from a recent breach of Mercor, an AI training startup that works with developers, and "made an educated guess about the model's online location based on knowledge about the format Anthropic has used for other models"--a phrase that many observers have speculated refers to a web URL--according to Bloomberg, which broke the story.
The perimeter failed. The model is fine, but its gate was a guessable URL. The irony is sharp.
Anthropic built a sophisticated AI to find the weakest link in any system. The weakest link in theirs, Bloomberg reports, was an old, dumb secret. The lesson isn't about a rogue AI.
It's about the human problem of hiding a key.
Related Reading
Browse all Anthropic news →
Research & Benchmarks Google's FACTS benchmark shows 70% factuality ceiling across four tests
Research & Benchmarks Databricks finds multi-step agents beat single-turn RAG by 21% to 38% on STaRK
Research & Benchmarks Nvidia's DLSS 4.5 beta adds 6x Multi Frame Generation for RTX 50 GPUs
Related Topic Trump cracks down on Anthropic after Amazon tip; staff largely foreign
Open Source
Related Topic Anthropic says Claude writes >90% of its code; AI pause button urged
LLMs & Generative AI
From Archives Google DeepMind's Vision Banana Outperforms SAM 3 and Depth Anything V3
From Archives DeepMind spinoff’s AI‑designed drugs enter human trials after AlphaFold 3
From Archives OpenAI's 'Spud' Beats Claude; April 30 Webinar on Agentspan 4‑Layer Production
LLMs & Generative AI
From Archives Claude adds direct connectors for Spotify, Uber Eats, TurboTax; mobile beta
LLMs & Generative AI
Common Questions Answered
How did Discord users gain unauthorized access to Anthropic's Mythos AI tool?
Discord users discovered relatively simple methods to bypass Anthropic's access restrictions without complex hacking techniques. The group of amateur sleuths managed to gain entry to the preview version of the Mythos AI model, which Anthropic had intentionally kept under tight controls due to its powerful security vulnerability detection capabilities.
Why did Anthropic describe the Mythos AI model as 'dangerously capable'?
Anthropic characterized Mythos as 'dangerously capable' because of its advanced ability to rapidly identify security vulnerabilities in software and networks. The model's potential for detecting critical security flaws was so significant that Anthropic initially limited access to only a handful of vetted partners to maintain control over its deployment.
What practical value has Mythos demonstrated in cybersecurity testing?
During Mozilla's early access testing, the Mythos AI tool helped identify 271 fixes for Firefox 150, demonstrating its substantial potential in proactively discovering and addressing software vulnerabilities. The incident highlights the tool's practical applications in cybersecurity while simultaneously raising questions about access control and potential misuse.
Further Reading
- Unauthorized Users Reportedly Slip Into Anthropic's Mythos AI System — National CIO Review
- Unauthorized users gained access to Anthropic's restricted Mythos AI model — The Next Web
- Unauthorized group has gained access to Anthropic's exclusive cyber tool Mythos — TechCrunch
- Anthropic's Mythos Accessed by Unauthorized Users — Bloomberg Intelligence (YouTube)
- AI Scare: Unauthorized Users Gain Access to Anthropic's Mythos — Bloomberg (YouTube)
