"Once you get to a minute, what can you do realistically? We need to start thinking beyond patching to address vulnerabilities," said Dick O’Brien, principal intelligence analyst at Symantec.
IRISSCON 2026: a blunt message from Symantec in Dublin
Speaking at IRISSCON 2026 in Dublin on November 11, Dick O’Brien delivered a concise, practical warning: organizations “can’t patch their way out of AI.” He argued that the security calculus has shifted because attackers are using AI not merely as a tool but as a force-multiplier that compresses the window between disclosure and exploitation.
Zero Day Clock: time-to-exploit collapsing from weeks to minutes
O’Brien pointed to the Zero Day Clock, a live dashboard that tracks time-to-exploit (TTE) measured from CVE disclosure to confirmed exploitation. The initiative estimates the average TTE fell from one week in 2025 to one day in 2026, and it projects an average of one minute in 2027. That trajectory — from days to minutes — is central to O’Brien’s argument that patching alone is no longer a viable defensive posture.
Audit-ready is a season. It shouldn't be.
Evidence in spreadsheets, controls drifting between audits, frameworks multiplying on flat headcount. Nubivance runs continuous compliance on Rapid7 Cyber GRC - SOC 2, HIPAA, ISO 27001, PCI, CMMC.
End the scrambleProof in the wild: PHP2Shell and pre-publication scans
The change is already visible in recent events. O’Brien cited the October 2026 PHP2Shell vulnerability, where the first reported exploitation attempts occurred just 19 minutes after the bug was made public. He also noted Amazon reported detecting an attempted exploitation of PHP2Shell four days before the vulnerabilities were published — a reminder that attackers and defenders can operate on different timelines, and that discovery and reconnaissance are happening prior to public disclosure.
Agentic AI orchestration and the DIVD incident
Beyond accelerating discovery, O’Brien warned of a shift toward “end-to-end agentic orchestration of attacks” within the next year — AI agents that carry out post-compromise tasks independently and at machine speed. He cited a concrete example: an attack on the Dutch Institute of Vulnerability Disclosure (DIVD) in September 2026 where, according to Symantec’s analysis of DIVD’s assessment report, AI agents executed post-compromise tasks autonomously. In that incident it took 19 minutes for detection, but 19 hours passed before a response began and an additional nine hours before the threat was contained — “This is what we’re dealing with. It’s an excess of 24 hours,” O’Brien said. The gap between machine-speed attack activity and human response windows is a core part of his case for rethinking defense.
Mitigation steps and what this means for security teams, procurement leaders, and policymakers
O’Brien urged an “assume exploit” mindset and a shift in metrics: “Stop measuring your mean time to patch and start measuring your mean time to mitigate. Assume an exploit is going to happen and the plan your preparedness for that,” he advised. He listed immediate mitigation tasks organizations can start now:
- Behavior blocking
- Default-deny for what runs
- Least privilege for every agent and token
- Inspect data before it leaves the device
- Bind every agent action to a named human
For longer-term resilience, O’Brien said organizations must integrate detection and protection technologies across multiple control points to detect behavioral anomalies of rogue or malicious AI agents.
How this lands for named groups in the discussion:
- Security teams: Shift measurement from patch speed to mitigation speed, implement the five immediate controls O’Brien listed, and prepare for post-compromise actions that can unfold in minutes rather than days.
- Procurement leaders and affected enterprises: Expect requirements to change toward vendor support for integrated detection/protection across control points and for mechanisms that enable least-privilege and binding agent actions to named humans.
- Policymakers and regulators: The call to “bind every agent action to a named human” and the rise of agentic orchestration highlight questions of accountability and traceability that will be practical focal points as organizations adopt mitigation-first postures.
O’Brien’s message was practical, not fatalistic: patching remains necessary, but it is no longer sufficient. With the Zero Day Clock projecting an average TTE of one minute in 2027 and real-world exploits already occurring within minutes, the defensive playbook must expand to include rapid mitigation, behavioral controls, and architectural changes that assume compromise rather than celebrate prevention alone.
https://www.infosecurity-magazine.com/news/shift-patching-mitigation-ai-era/
