Anthropic Unveils Claude Mythos Preview With Powerful Zero-Day Detection Capabilities

In Cybersecurity News - Original News Source is cybersecuritynews.com by Blog Writer

Anthropic has introduced Claude Mythos Preview, an advanced language model with extraordinary capabilities for discovering and autonomously exploiting undiscovered zero-day vulnerabilities.

To ensure these powerful tools are used defensively, the company has launched Project Glasswing to collaborate with industry partners and patch critical software systems.

Claude Mythos Preview represents a massive upgrade over older models like Opus 4.6, which could find bugs but struggled to turn them into working exploits.

During internal tests using open-source software, the new model successfully achieved full control-flow hijacking on 10 fully patched targets.

These advanced offensive skills were not explicitly programmed; rather, they emerged naturally from the model’s overall improvements in logical reasoning and autonomous coding.

Autonomous Exploit Generation

The model can autonomously chain together multiple software flaws to create highly complex attacks that bypass modern security boundaries.

For example, it successfully wrote web browser exploits that evaded strict sandboxes and bypassed kernel address space layout randomization (KASLR) to gain elevated privileges.

Mythos Preview leads by a wide margin in exploit success (source: Anthropic)

Because the tool is highly automated, even users without any formal cybersecurity training have used it to generate fully working remote code execution exploits overnight.

When unleashed on real-world software, the AI agent discovered critical zero-day bugs that had remained hidden from human researchers for decades.

It successfully identified a 27-year-old memory corruption vulnerability in OpenBSD. This operating system is widely respected for its rigorous security standards.

Furthermore, it found a 16-year-old flaw in the highly audited FFmpeg media library by analyzing how the software decodes specific video frames.

The OpenBSD vulnerability was caused by a complex signed integer overflow in the network transmission control protocol, which the AI used to trigger a system crash.

The FFmpeg bug occurred due to a mismatch in integer sizes and memory initialization, allowing an attacker to force the program to write out-of-bounds data.

To find these flaws, the AI operates inside an isolated testing environment where it reads source code, tests hypotheses, and writes proof-of-concept exploits completely on its own.

Anthropic acknowledges that releasing such a powerful vulnerability-discovery tool could temporarily give malicious hackers a dangerous advantage.

To prevent this, Project Glasswing limits initial access to trusted defenders who can use the model to fix deep-seated bugs before they are actively exploited in the wild.

Ultimately, security experts believe that as the industry adapts, these advanced AI models will become essential defensive tools, making the global software ecosystem significantly safer.

Follow us on Google News, LinkedIn, and X for daily cybersecurity updates. Contact us to feature your stories.