Sophos’s Exploit Path Verification (EPV) will leverage OpenAI’s GPT cyber models via the Daybreak Defense Network to evaluate asset state, patch status, endpoint protection policies, network reachability, identity and privilege information, and known exploit availability, and to reason over chained paths from multiple lower-severity findings.

Sophos serves more than 625,000 organizations worldwide, including 40,000 MDR customers, supported by a large partner ecosystem for delivering security solutions.

Sophos security analysts will review EPV results to ensure accuracy and proper context.

Sophos CTO John Peterson emphasizes that EPV clarifies what in an environment is reachable by attackers and prioritizes fixes, all supported by evidence.

EPV aims to determine whether a flaw is reachable and exploitable in a given environment, addressing the gap between generic vulnerability severity and actual risk.

EPV is targeted at enterprise and mid-market customers within Sophos Managed Risk, with availability and early-access details to be announced later.

EPV is being integrated into Sophos Managed Risk to prioritize and manage exploitable vulnerabilities by identifying truly exploitable weaknesses in an environment.

The initiative builds on Sophos’s collaboration with OpenAI through the Daybreak Defense Network, begun in June 2026, combining frontier AI reasoning with environment-specific data and human oversight.

EPV will assess whether a control blocks a technique class or only a public PoC and will draft remediation text ready for ticketing.

The capability will identify attack paths formed by chaining multiple lower-severity vulnerabilities and will provide remediation recommendations.

Findings will be categorized as Confirmed Exploitable, Blocked by a Control, Not Reachable, or Insufficient Evidence, with supporting evidence for each verdict.

OpenAI executives emphasize responsible use of frontier AI with guardrails in the Daybreak Defense Network, applying frontier reasoning to defensive problems.