BCG: 35% of Companies Already Hit by a Significant AI-Enabled Attack

More than a third of organizations — 35% — report a significant impact from an AI-enabled attack in just the past twelve months, according to BCG’s 2026 CISO Survey. Data leakage is the most cited AI-related risk, and the average organization now reports three significant breaches and 25 sensitive data incidents a year.

BCG’s argument is that this stopped being a delegable problem the moment AI expanded the attack surface faster than most governance structures adapted to cover it. The firm lays out five actions specifically for CEOs, not CISOs: prioritize critical assets and processes; build detection and recovery systems measured in minutes rather than days; govern third-party and AI-ecosystem dependencies; embed security into product design from the start, not after; and stand up cross-functional teams with clear decision rights spanning CISOs, CROs, CTOs, legal, procurement, HR, and business-unit leaders.

That decision-rights point is the one worth sitting with. Palo Alto Networks’ own numbers, reported the same week on the vendor side, show why: Unit 42 researchers simulated a full ransomware campaign — initial entry through data exfiltration — in 25 minutes, against a typical enterprise detection window measured in days. Palo Alto’s answer has been to consolidate its stack across five platform pillars built through 20-plus acquisitions rather than sell point products, a bet that’s helped push the company past an $11.4 billion revenue run rate and a market cap near $265 billion. The vendor-side incentive and the CEO-side governance argument point at the same underlying fact: a security team working from fragmented tools and unclear ownership can’t close a 25-minute attack window, no matter how good any single tool is.

For leadership teams treating AI security as a technical delegation, BCG’s numbers are the corrective — the exposure is already organizational, and the fix has to be too.