2026-08-18 · Source: Insurance Journal
Summary in 3 Points • Agentic AI research led to a 36% increase in new cyber vulnerabilities • Only a 10% rise in actively exploited vulnerabilities was reported by CISA • Beazley Security noted 67% of ransomware intrusions involved compromised credentials --- Agentic artificial intelligence is accelerating the discovery of cyber vulnerabilities, with Beazley Security reporting a 36% increase in newly disclosed vulnerabilities in Q2 2026. Despite this rise, the increase in vulnerabilities actively exploited and added to CISA's Known Exploited Vulnerabilities catalog was only 10%. Beazley Security highlighted that attackers are still primarily using traditional methods like compromised credentials, which accounted for 67% of ransomware intrusions investigated. The report also noted a significant number of new high-risk vulnerabilities, with 5,600 identified as potentially harmful if unpatched.
The increase in discovered vulnerabilities due to agentic AI could lead to heightened scrutiny in underwriting cyber insurance policies. Insurers may need to reassess the risk exposure of clients, especially those with inadequate cybersecurity measures. The reliance on compromised credentials for ransomware attacks suggests a potential area for insurers to focus on in policy wording and coverage limits. As AI continues to evolve, the London Market might consider investing in AI-driven tools to better assess and mitigate cyber risks.