OpenAI Launches Restricted Cybersecurity Model Amid Rising AI Security Concerns
OpenAI has unveiled GPT 5.4 Cyber, a specialized AI model designed for cybersecurity defense, featuring fewer restrictions on security-related queries for verified professionals. This launch occurs just days after rival Anthropic released Claude Mythos Preview, a powerful model capable of identifying critical software vulnerabilities, which raised significant concerns about potential misuse in cyberattacks. To mitigate risks, OpenAI is limiting access to GPT 5.4 Cyber through its Trusted Access for Cyber programme, available only to vetted security vendors, organizations, and researchers. The new model includes advanced capabilities such as binary reverse engineering, allowing experts to analyze compiled software for malware without source code access. This strategic move follows Anthropic's decision to restrict its own model to twelve founding partners, including major tech firms like Amazon, Apple, and Microsoft, due to the unprecedented cybersecurity risks identified by its engineers. Both companies are navigating the delicate balance between leveraging AI for defensive security improvements and preventing the technology from enabling sophisticated cyberattacks. The simultaneous releases highlight the intensifying competition and regulatory caution within the AI industry regarding high-stakes security applications.
Wire timeline
OpenAI Launches Restricted Cybersecurity Model Amid Rising AI Security Concerns
OpenAI has unveiled GPT 5.4 Cyber, a specialized AI model designed for cybersecurity defense, featuring fewer restrictions on security-related queries for verified professionals. This launch occurs just days after rival Anthropic released Claude Mythos Preview, a powerful model capable of identifying critical software vulnerabilities, which raised significant concerns about potential misuse in cyberattacks. To mitigate risks, OpenAI is limiting access to GPT 5.4 Cyber through its Trusted Access for Cyber programme, available only to vetted security vendors, organizations, and researchers. The new model includes advanced capabilities such as binary reverse engineering, allowing experts to analyze compiled software for malware without source code access. This strategic move follows Anthropic's decision to restrict its own model to twelve founding partners, including major tech firms like Amazon, Apple, and Microsoft, due to the unprecedented cybersecurity risks identified by its engineers. Both companies are navigating the delicate balance between leveraging AI for defensive security improvements and preventing the technology from enabling sophisticated cyberattacks. The simultaneous releases highlight the intensifying competition and regulatory caution within the AI industry regarding high-stakes security applications.
euronews