Story
Cybersecurity experts at Harvard's Berkman Klein Center are urging government and business leaders to regulate agentic AI systems before they are exploited by bad actors. The call comes as IBM data shows a 44% year-over-year increase in cyberattacks targeting public-facing software and applications, many utilizing AI. High-profile incidents include the November 2025 data breach of Anthropic, where attackers used AI models to scan for weaknesses in source code. Experts warn that AI-powered phishing attacks have become highly sophisticated, removing traditional telltale signs like misspellings. Panelists including former White House cyber director Robert Knake advocate for a safe harbor framework that rewards companies implementing basic security measures while holding those negligent liable for breaches. Harvard Gazette reported the core details. IBM Threat Intelligence noted additional context. Harvard Berkman Klein Center added corroborating details.