Bypassing safety controls on AI models poses immediate risk to global security

Microsoft co-founder Bill Gates warned that malicious actors will likely increase AI-powered cyberattacks over the next 12 months, citing the removal of safety controls from advanced models as a primary risk. Speaking on Sunday, 11 October, Gates stated that the immediate threat of AI misuse outweighs longer-term concerns regarding autonomous systems spiralling out of control. During an interview with ABC News, the billionaire philanthropist identified cybersecurity as the sector most vulnerable to AI exploitation in the short term. He noted that the primary danger arises when safeguards built into advanced AI models are bypassed or removed, allowing individuals to weaponize the technology for digital warfare. Malicious exploitation of unregulated models Gates indicated that unrestricted access to the latest large language models (LLMs) could result in \"negative surprises\" by the end of 2027. While AI offers defensive capabilities for security teams, the ease with which these tools can generate sophisticated phishing campaigns, automate malware creation, and find software vulnerabilities has lowered the barrier for cybercriminals. The warning comes as global regulators debate the level of transparency required for open-source AI models. Proponents of strict regulation argue that releasing powerful model weights without oversight allows bad actors to strip away safety filters designed to prevent the generation of harmful code or instructions. Immediate impact on global…