Anthropic’s latest AI model could let hackers carry out attacks faster than ever. It wants companies to put up defenses first

In a significant move to combat the escalating threat of cyberattacks, Anthropic has announced that its latest AI model, dubbed Mythos, will be made available to several major tech and cybersecurity companies. This initiative aims to help these firms strengthen their defenses against potential vulnerabilities that could be exploited by malicious actors. Notable companies such as Amazon, Apple, Cisco, Google, JPMorgan Chase, and Microsoft will have access to this cutting-edge AI tool, enabling them to identify bugs in their software and test the effectiveness of various hacking techniques on their products. However, the public launch of Mythos is still on hold due to concerns about its potential misuse by cybercriminals. Logan Graham, who leads the team responsible for the AI models’ defenses at Anthropic, expressed caution, stating, "We did not feel comfortable releasing this generally. We think that there’s a long way to go to have the appropriate safeguards in place." This caution reflects a growing unease in both Washington and Silicon Valley regarding AI's impact on cybersecurity. Experts have pointed out that AI agents like Mythos can search for vulnerabilities at a speed and scale that far exceeds human capabilities, fundamentally changing the landscape of cybersecurity. A single AI agent could potentially identify and exploit vulnerabilities more rapidly and persistently than hundreds of human hackers combined. In light of this, Anthropic has briefed senior officials across the U.S. government about Mythos' capabilities, both offensive and defensive, and has committed to supporting governmental testing and evaluation of the technology. Anthropic aims for the selective release of Mythos to major companies, which collectively serve billions of users, to help level the playing field against cyber attackers. Their objective is to identify and address critical security flaws in widely used software before these vulnerabilities can be publicly exploited. Other organizations, including chipmakers Broadcom and Nvidia, the Linux Foundation, and cybersecurity firms like CrowdStrike and Palo Alto Networks, will also gain access to Mythos. As AI's role in cybersecurity expands, Graham emphasized the urgency for defenders to adapt quickly. He remarked, "The world is very different now if these model capabilities are going to be in our lives." A recent blog post hinted at Mythos' advanced capabilities, suggesting it is significantly ahead of its competitors in identifying and exploiting vulnerabilities. While some concerns surrounding Mythos' potential for misuse have been deemed exaggerated by experts, the underlying truth remains: unless there is a strategic shift, the gap between cyber attackers and defenders could continue to widen. Anthropic asserts that Mythos has already uncovered thousands of previously unknown software vulnerabilities, a rate that far surpasses that of human researchers. Such findings are invaluable to both cybercriminals and espionage agencies aiming for stealthy hacks. However, it's crucial to note that cybersecurity professionals have been leveraging AI for protective measures long before the arrival of Mythos. Gadi Evron, a security researcher, emphasized the disparity in AI capabilities between attackers and defenders, stating that defenders must harness AI effectively to keep pace with evolving threats.