AI Is Finding Flaws Faster Than Anyone Can Fix Them: Can Organisations Even Hear the Warnings?
Artificial intelligence models now find software vulnerabilities faster than organisations can fix them. According to HackerOne's platform data, the backlog of valid vulnerability reports is up 131% over the last two years, while critical and high-severity findings are rising and remediation falls further behind. The author argues that finding flaws is no longer the hard part. The harder question is whether anyone is listening when an outsider reports one.
The piece, published by CyberScoop, focuses on the US Task Force on Super Intelligence. Its charter requires a report on how superintelligence affects the balance between offence and defence in cybersecurity, and an assessment of preparedness across healthcare, communications, banking, energy, water, transportation and emergency services. The author warns that if these assessments ignore third-party testing and disclosure, they will start from an incomplete picture. Internal teams often share the same assumptions as the people who built a system, and may miss what an outsider spots quickly.
The author, who co-wrote a paper on third-party AI flaw disclosure with researchers from Stanford, Princeton, MIT and elsewhere, says whether an organisation accepts, triages and acts on outside reports is one of the best indicators of defensive maturity. The US government has already moved this way: the Office of Management and Budget has required agencies to accept outside reports since 2020, a June executive order extends the expectation to covered federal contractors, and a bipartisan bill in the House defence authorisation act points the same direction.
The author recommends that the task force, within its 120-day mandate, benchmark how quickly reports are triaged and fixed, map where independent reporting is unavailable across critical sectors, extend standardised flaw reporting to AI systems, and recommend legal safe harbours for good-faith researchers.
Why it matters
Many organisations treat internal testing and the existence of a reporting channel as proof of readiness, but this piece argues that neither shows whether outside warnings are actually triaged and fixed. With AI-driven discovery growing the report backlog by 131%, the weak point shifts from finding flaws to handling them, which hits small operators such as rural hospitals, community banks and local utilities hardest. Leaders should measure time to triage and fix, not just whether a channel exists.
What to do
- Publish a vulnerability disclosure contact and policy so outside researchers know how to report a flaw.
- Track how long reported vulnerabilities take to be triaged and fixed, and review the backlog regularly.
- Assign a named owner to handle incoming outside reports and drive them to remediation.
- Prioritise critical and high-severity findings in your remediation queue.