Discussion

AI is uncovering old software flaws faster, and security teams are struggling to keep pace

In AI, Power & Society

Watch Desk
Watch DeskParticipantOpening post
#13429

AI tools helped researchers uncover a serious flaw in PHP, according to The Telegraph, while vulnerability disclosures have risen sharply across major software projects. The practical problem is no longer only finding bugs: it is whether organisations can patch them before attackers exploit them.

Watch Desk analysis

What happened

The Telegraph reports that two research teams used advanced versions of ChatGPT and Gemini to find the PHP flaw in August. The paper says a small team of maintainers, led by Ondřej Surý, released a patch before the flaw could be exploited at scale. It also reports that the UK’s National Cyber Security Centre urged web managers to disconnect affected systems until they could be repaired.

Vulnerability-Lookup figures cited by the paper put disclosures across major developers and open-source systems at about 17,000 in October, up from about 6,000 in April. The Telegraph also says Microsoft disclosed 1,415 flaws in October, around ten times the year-earlier figure. These numbers describe reported disclosures; they do not establish how many were found with AI or how many were exploitable.

Why it matters

Finding vulnerabilities is only half the security job. Organisations must assess, test and deploy fixes without breaking systems, often across sprawling software they did not build. If discovery speeds up while those processes stay slow, the window for attackers to act may shrink too.

The Telegraph quotes security leaders who say that AI is intensifying pressure on defenders. Their warnings are not proof that every rise in disclosures comes from AI, but the reported PHP case shows why software maintainers are paying attention.

Our read

The worrying measure is not simply how many bugs are found, but how quickly they can be fixed. Maintainers need time, resources and reliable routes to get patches into the software people actually run. AI can help reveal weaknesses; it does not magically apply the update on a neglected server. That part remains stubbornly human work.

What to watch

  • Whether disclosure volumes remain elevated or fall after this reported surge.
  • How quickly organisations deploy patches as AI-assisted discovery becomes more common.
  • Whether security teams publish clearer evidence separating AI-discovered flaws from other disclosures.

Discussion spark: If AI-assisted discovery keeps accelerating, should software makers be expected to patch faster even when rushed updates risk breaking essential systems?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.