New Delhi : OpenAI has raised a sharp alert about the direction its next generation of AI models is heading in, calling out the possibility of “high” cybersecurity risks as their capabilities expand. The warning lands at a time when the company is under pressure to move faster than Google’s Gemini project, and is internally pushing teams to deliver more powerful versions of ChatGPT.
In a detailed blog post published on Wednesday, the company acknowledged that the pace of progress could soon reach a point where its models are capable of tasks that were previously in the domain of advanced hacking groups. OpenAI noted that future systems might be able to create working zero-day remote exploits or even assist with high-level intrusion operations targeting industrial network attacks that historically required specialised human expertise.
Despite the red flags, the company said the intention is to ensure the models’ strengths tilt towards defensive roles rather than offensive misuse. OpenAI wrote that as the technology grows more potent, it is “investing in strengthening models for defensive cybersecurity tasks and creating tools that enable defenders to more easily perform workflows such as auditing code and patching vulnerabilities”. The effort is part of a larger attempt to improve the models in ways that benefit cybersecurity teams instead of putting them at risk.
To support that, OpenAI plans to rely on layered safeguards that include stricter access controls, hardened infrastructure, egress monitoring and several internal checks. The Microsoft-backed company also revealed upcoming plans for a program that would provide tiered access to enhanced capabilities for vetted researchers and cyberdefence groups. The idea is to give security professionals early access to advanced AI tools, but only after meeting strict qualification standards.
The company is also setting up a new advisory group called the Frontier Risk Council, which will bring seasoned cyber defenders and security experts into close collaboration with OpenAI’s technical and policy teams. This group will initially focus on cybersecurity but is expected to expand to cover other high-risk areas associated with frontier AI models.
While the public messaging focuses on risk management and responsible scaling, internal developments paint a slightly different picture of the urgency at OpenAI. Reports suggest that CEO Sam Altman has asked employees to accelerate progress by leaning more heavily on user-generated inputs, including one-click feedback from ChatGPT users rather than relying mainly on trained evaluators. This change is seen as controversial because it increases the weight of real-time, unfiltered user data in model training, which can introduce inconsistencies but also allows the system to learn much faster.
