OpenAI’s release of GPT-5.4 has raised significant cybersecurity concerns as it is the first general-purpose AI rated as a “high” cybersecurity threat. The model demonstrated its capabilities by scoring an impressive 88% on professional-level “Capture the Flag” challenges, which test its ability to crack encryption and exploit vulnerabilities in simulated networks. This high performance indicates that GPT-5.4 can independently plan and execute cyberattacks, automating the discovery of system flaws without human intervention, thus posing a rapid scaling threat for cybercriminals. OpenAI has implemented safety mitigations, including real-time monitoring and tiered access, to address the risks associated with this powerful technology.
OpenAI’s GPT-5.4 rated high cybersecurity threat after capture the flag tests
