Home›Insurance›Industry & Deals›Agentic AI research boosts disclosed vulnerabilities,…

Agentic AI research boosts disclosed vulnerabilities, but exploits lag

Beazley Security said ransomware intrusions investigated in Q2 were driven by compromised credentials, which accounted for 67% of cases.

Reported · Written up by our desk · Industry & Deals

Agentic artificial intelligence is accelerating the discovery of new cyber weaknesses, but it has not yet changed how many of those vulnerabilities get actively exploited, according to research from Beazley Security.

Beazley Security’s Q2 2026 Quarterly Threat Report found that agentic AI used in vulnerability research drove a 36% quarter over quarter increase in newly disclosed vulnerabilities. In contrast, vulnerabilities confirmed as actively exploited and added to the CISA Known Exploited Vulnerabilities catalog rose only 10% during the same period.

The report also pointed to a rising overall risk picture, with more than 20,700 new vulnerabilities publicly disclosed in Q2. Of that total, nearly 5,600 were considered high risk, meaning they could be exploited remotely if left unpatched, and Beazley Security issued advisories for 21 of those high risk vulnerabilities.

While Beazley Security said attackers are experimenting with AI, it said they are still relying largely on established tactics, including compromised credentials, which accounted for 67% of the ransomware intrusions the firm investigated during Q2. The firm added that organizations should remain focused on cybersecurity hygiene and consider assessments to determine how AI is used within their enterprises.

More like this

Sources