Anthropic has published a preliminary report for Project Glasswing, a cybersecurity initiative launched in April that aims to prevent AI cyberattacks with AI. The initiative is powered by the company’s unreleased model Cloud Mythos Preview, which Anthropic says has helped its partners find more than ten thousand vulnerabilities in total just one month after Glasswing’s launch. Additionally, it says most of its partners have “found hundreds of critical or high-severity vulnerabilities in their software” using the model.
The company said its partners’ bug finding rate has increased more than tenfold. Cloudflare found 2,000 bugs, 400 of which are at high severity or critical. Mozilla previously reported that it found and fixed 271 vulnerabilities in Firefox, which was 10 times more than the number of vulnerabilities found in older versions of the browser using other cloud models.
Microsoft’s recent announcement that its patch releases “will be trending in a big way for some time” is apparently due to bugs found through the Mythos preview. Anthropic also used Mythos Preview to scan 1,000 open-source projects over the past few months and found 6,202 high- and critical-severity vulnerabilities out of 23,019. Although the company has not included it in the report, a security research firm recently claimed that it has found a way to break into macOS, an operating system known for tight security, with the help of Mythos’ bug-discovery capabilities.
The company explained in its report that it has not yet released the Mythos Preview to the public, because no company (including itself) has developed security measures strong enough to prevent misuse of models like it. However, it intends to release a “Mythos-class model” in the future when these security measures become available. Currently, the project plans to work with partners such as the US and other governments to expand the availability of Glasswing. This indicates that the company is on its way to improving its relations with the US government. The company is already working with several partners at the moment, including Amazon Web Services, Apple, CrowdStrike, Google, JPMorganChase, NVIDIA, and Palo Alto Networks, in addition to those we have already mentioned.
Anthropic is reportedly set to become profitable for the first time since its founding in 2021. According to a recent report by wall street journalIt is on track to report revenue of $10.9 billion with an operating profit of $559 million for the quarter ending in June. However, the company is not expected to remain profitable in the coming quarters, as it intends to invest more money in computing resources and other expenses.
<a href