US EditionMove over Bitcoin and Quantum Risks. Anthropic's Mythos AI Changes Everything for DeFi
Anthropic said Mythos Preview found working exploits in 72.4% of tests and is sharing it with 40 partners to harden defenses.
- On Tuesday, Anthropic announced its new Mythos Preview model but restricted public access, warning the AI can autonomously identify and exploit zero-day vulnerabilities in major operating systems and web browsers.
- Mythos Preview represents a significant leap in capability, achieving a 72.4 percent exploit development success rate compared to Claude Opus 4.6, which Anthropic noted was not skilled at creating working exploit code.
- "AI models have reached a level of coding capability" where they surpass most humans at exploiting software vulnerabilities, Anthropic said, noting the model can autonomously write complex code to escape OS sandboxes.
- Around 40 organizations, including Amazon Web Services, Google, and Microsoft, joined Anthropic's Project Glasswing to test their systems using Mythos, subsidized by $100M in usage credits and $4M in direct security donations.
- Researchers are responsibly disclosing thousands of critical vulnerabilities identified by Mythos to help organizations defend against these advanced threats before adversaries exploit them, noting the model is "dense and technical.
27 Articles
27 Articles
Op-Ed: We've Hit a Scary 'Turning Point' for AI
Those worried about advances in artificial intelligence should know an alarming "turning point" has already arrived sooner than expected, and with "profound geopolitical implications," according to New York Times columnist Thomas L. Friedman. Anthropic's new model, Claude Mythos Preview, can not only write software code better than any model before...
Move over bitcoin and quantum risks. Anthropic's Mythos AI changes everything for DeFi
Claude Mythos Preview identified thousands of zero-day vulnerabilities across every major operating system and browser, including in cryptography libraries that DeFi infrastructure depends on.
In the course of internal testing, thousands of computer flaws have been identified in the new d的IA d的Anthropic, Mythos model. Its release is therefore postponedThe start-up of artificial intelligence (IA) Anthropic announced on Tuesday that it is postponing the marketing of its new d的IA model, Mythos, the time to fill thousands of computer flaws identified by the interface.According to the Californian company, during in-house testing, Mythos ha…
Meet Claude Mythos, Anthropic AI so powerful it can hack any software, preview out now
Anthropic has a new AI model, called Claude Mythos Preview, which is so good that the AI startup has no plans of making it available to the public anytime soon. Instead, Claude Mythos Preview will be made available to a consortium of over 40 tech companies, such as Apple and Google, that will use the model to fix potential cybersecurity vulnerabilities.
Coverage Details
Bias Distribution
- 50% of the sources lean Left
Factuality
To view factuality data please Upgrade to Premium
