Modern AI systems are effectively universal advisers that help people perform harmful actions, according to security technologist Bruce Schneier. Writing in a recent opinion piece, Schneier argues that AI is rapidly widening the gap between skill and ability, enabling individuals with minimal expertise to carry out sophisticated cyber-attacks.
Five Eyes Warning on AI Cyber Risks
Earlier this week, national security agencies from the Five Eyes intelligence alliance—comprising Australia, Canada, New Zealand, the United Kingdom, and the United States—jointly released a statement warning of increasing cyber risks from AI models. The statement highlighted AI's ability to autonomously hack into systems and networks. While the advice echoed longstanding security recommendations, the agencies stressed new urgency due to AI's rapid evolution.
Schneier notes that cyber-attacks have been a significant issue long before generative AI, but the technology is accelerating a decoupling of skill and ability. Historically synonymous, these terms are now diverging as computers empower users to accomplish more—including more damage—with less direct expertise.
From Hacker Elites to AI-Powered Threats
In 1998, seven members of the hacker group L0pht testified before Congress, claiming they could take down the internet in 30 minutes. Schneier contrasts their deep technical skill with so-called “script kiddies,” who use prewritten tools with minimal understanding. AI now allows even less skilled individuals to launch automated attacks with minimal prompting.
“The thing about people with ability but no skill is that they are often outsiders, not part of any professional community, and not bound by any rules or norms,” Schneier writes. He draws parallels to doctors who know how to poison or engineers who know how to demolish bridges, noting that professional training instills ethical codes that casual AI users lack.
Guardrails and Open-Source Models
Current AI companies attempt to build guardrails preventing harmful queries, but Schneier argues this approach is unsustainable. Smaller, cheaper, open-source models—including those running locally on personal computers—match the capabilities of frontier models from OpenAI and Anthropic. These models can be shared freely without restrictions, much like script kiddie tools of the past.
Instructing AI models to monitor and report malicious prompts also fails, as open-source versions will not comply. Schneier estimates such measures might buy “a few months at best.” Making models inherently unable to cause harm is equally impractical, because the same knowledge required for defense—such as finding vulnerabilities in code—can be used for attack.
Defense Through AI
The Five Eyes statement recommends using AI to strengthen cybersecurity defenses: detecting vulnerabilities earlier, improving software quality, monitoring unusual behavior, and responding faster to incidents. Schneier endorses this approach, urging similar AI-driven defense across all risks heightened by the technology.
“The rapid pace of frontier AI development means cyber risk assumptions can become outdated in months, not years,” the Five Eyes noted. Schneier concludes that super-powered humans with AI assistants will create a world of increased volatility, capable of both wonderful and horrible outcomes.
