AI giants compete fiercely! Mythos quickly aims to take the lead as OpenAI speeds up the launch of a new cybersecurity model.

OpenAI launched a derivative version of its latest artificial intelligence model GPT-5.5-Cyber on Thursday, currently open for use in a limited preview format to vetted cybersecurity teams. Just a month ago, competitor Anthropic's preview version of Claude Mythos attracted significant attention from investors and U.S. government officials. OpenAI stated in their official blog that the preview version of GPT-5.5-Cyber is not intended to significantly enhance networking capabilities but rather relaxes the execution restrictions for security-related tasks through targeted training. The company had released the general version of GPT-5.5 at the end of last month. OpenAI mentioned that this cybersecurity-specific model allows compliance teams to easily utilize their latest large models for vulnerability identification and classification, patch effectiveness verification, malware analysis, and other workflow processes. The built-in security mechanisms in the general version of GPT-5.5 will greatly increase the operational difficulty of such specialized tasks. In their blog post, OpenAI stated that GPT-5.5-Cyber is open to a few select partners to facilitate the study of advanced security workflows, adapting to professional scenarios with special requirements for exclusive access. At the time of the release of the Mythos model by Anthropic last month, they also initiated a network security special program called "Project Glasswing," which grants model permissions to only a few selected companies. Anthropic's CEO Dario Amodei met with senior Trump administration officials to discuss the technical capabilities and potential impact of the model, despite the company being recently blacklisted by the Pentagon for "supply chain risks." Reportedly, the U.S. Department of Defense has requested Anthropic to allow unrestricted use of their model for all legal purposes, while Anthropic hopes to ensure that their technology is not used for fully automatic weapons or mass surveillance. Early testing information about Mythos caused market turbulence. In late March of this year, as the potential risks of the model autonomously discovering vulnerabilities and bypassing existing network defense systems gradually surfaced, market risk aversion quickly escalated, putting pressure on the cybersecurity sector with stocks like CrowdStrike, Palo Alto Networks, and Zscaler experiencing significant declines. Mythos also raised high alarms within U.S. regulatory agencies. It was reported that last month, Federal Reserve Chairman Jerome Powell and Treasury Secretary Scott Bennett held a closed-door meeting with Wall Street bank executives to discuss the potential network security threats posed by Mythos. Additionally, Vice President Vance and Bennett also had advanced phone discussions with major tech company executives. It is worth noting that the competition between Anthropic and OpenAI has intensified. Anthropic, which originally focused on enterprise AI, has recently aggressively entered the consumer market with several practical products gaining rapid popularity. Leveraging a high-profile ad during the Super Bowl, they positioned Claude as an ad-free alternative to ChatGPT, directly challenging OpenAI. This, combined with the discussions regarding U.S. Pentagon's relationship, has boosted the popularity of Claude, securing the second position in the U.S. Apple App Store free chart, sandwiched between ChatGPT in first place and Google Gemini in third, further intensifying the strong impact on OpenAI.