"OpenAI states, "Our goal is to make these tools as widely available as possible while preventing misuse. We design mechanisms which avoid arbitrarily deciding who gets access for legitimate use and who doesn't.""
"OpenAI says its intention is to learn by putting the model into the world and improving it over time. "As we better understand both their capabilities and risks, we update our models and safety systems accordingly.""
OpenAI introduced GPT-5.4-Cyber, a model tailored for defensive cybersecurity, expanding access through its Trusted Access for Cyber program. This initiative aims to provide tools to thousands of verified defenders and groups safeguarding critical infrastructure. OpenAI emphasizes the importance of clear criteria for access and intends to learn from real-world usage to enhance the model's capabilities and safety. Security experts highlight the distinction between identifying and fixing bugs, stressing the ongoing evolution of AI in cybersecurity.
Read at Securitymagazine
Unable to calculate read time
Collection
[
|
...
]