AI restrictions hinder the work of cybersecurity professionals

With the development of AI technologies, major tech corporations are implementing strict guardrails to prevent their models from being used for malicious purposes. However, these barriers, intended to prevent cyberattacks, are now causing serious problems not only for cybercriminals but also for legitimate researchers and security engineers who protect systems. This is reported by Techcrunch.com reports.

In recent months, leading companies like OpenAI and Anthropic have launched special verification programs to keep their models secure. For example, Anthropic's Mythos and Fable models were placed under export controls by the US government due to concerns that they could be used to carry out cyberattacks. Although these restrictions were later partially eased, cybersecurity professionals argue that this approach is weakening defense systems.

The fine line between defense and attack

Cybersecurity researchers often need to think like hackers and use similar tools to identify and patch vulnerabilities in systems. According to TechCrunch, when many experts send requests to AI models to analyze code and find errors, the system evaluates this as a "dangerous action" and refuses to respond.

Chris Anley, Chief Scientist at NCC Group, explains that a command to "fix" software code can serve as a guide for both defense and attack. "The same tool acts as both an offensive and defensive weapon, and they cannot be separated. If an AI model refuses to analyze code, it makes the work of defenders more difficult," the expert says.

Subjective corporate decisions

Prominent security researcher Mark Dowd believes that private companies independently deciding what is safe and what is dangerous negatively impacts the development of the field. He notes that major tech giants are imposing arbitrary restrictions without considering the complex aspects of cybersecurity.

Currently, OpenAI, through its Trusted Access for Cyber, and Anthropic, through its Cyber Verification Program, only allow organizations that have passed special vetting to use models with reduced restrictions. However, such "closed-door" policies limit the capabilities of independent researchers.

Cybersecurity is becoming an increasingly relevant topic in Uzbekistan as well. Local experts facing such obstacles when using global AI models could slow down the process of protecting the country's information systems. After all, to counter modern cyberattacks, professionals must have full access to the most advanced intellectual tools.