Research from the University of Illinois

web

ibm.com·ibm.com/think/insights/chatgpt-4-exploits-87-percent-one-...

Key empirical evidence that frontier LLMs lower the barrier to cyberattacks; relevant to AI risk assessments, deployment policy debates, and discussions of capability thresholds for dangerous use.

Metadata

Importance: 72/100news articlenews

Summary

This IBM Think article summarizes University of Illinois research demonstrating that GPT-4 can autonomously exploit 87% of 'one-day' (recently disclosed but unpatched) cybersecurity vulnerabilities when given CVE descriptions. The finding highlights the dual-use risk of advanced LLMs as tools for automated cyberattacks, requiring only publicly available vulnerability information to achieve high exploitation rates.

Key Points

•GPT-4 successfully exploited 87% of one-day vulnerabilities in controlled tests when provided CVE descriptions.
•The research used real-world CVEs, showing LLMs can translate public vulnerability disclosures into working exploits.
•Weaker models (GPT-3.5, open-source LLMs) performed significantly worse, suggesting capability thresholds matter for cyber risk.
•Findings raise urgent questions about responsible disclosure timelines and LLM access controls in security contexts.
•Demonstrates that frontier AI models represent a qualitative leap in the accessibility of cyberattack capabilities.

Cited by 1 page

Page	Type	Quality
Cyberweapons Risk	Risk	91.0

Cached Content Preview

HTTP 200Fetched Mar 20, 202614 KB

[Skip to content](https://www.ibm.com/think/insights/chatgpt-4-exploits-87-percent-one-day-vulnerabilities#main-content)[IBM logo](https://www.ibm.com/)

Software

Infrastructure

[Consulting](https://www.ibm.com/consulting?lnk=L0G)Support

[Overview](https://www.ibm.com/mysupport/s/?language=en_US&lnk=flathl)[Community](https://community.ibm.com/community/user/community?lnk=flathl)[Developer](https://developer.ibm.com/?lnk=flathl)[Documentation](https://www.ibm.com/docs/en?lnk=flathl)[IBM Cloud platform](https://www.ibm.com/products/cloud/support?lnk=flathl)[Implementation](https://www.ibm.com/products/expertlabs?lnk=flatitem)[Training](https://www.ibm.com/training/?lnk=flathl)[Technology Lifecycle Services](https://www.ibm.com/services/technology-lifecycle-services?lnk=flathl)[Think 2026](https://www.ibm.com/events/think)

[Cart](https://www.ibm.com/store/en/us/checkout)

My IBM
Log in


[Think](https://www.ibm.com/think)

- [Artificial intelligence](https://www.ibm.com/think/artificial-intelligence)
- [Cloud](https://www.ibm.com/think/cloud)
- [Security](https://www.ibm.com/think/security)
- [News](https://www.ibm.com/think/news)
- Videos






  - [Overview](https://www.ibm.com/think/videos)
  - [AI Academy](https://www.ibm.com/think/videos/ai-academy)
  - [Think 2025 on demand](https://www.ibm.com/think/videos/think-keynotes)
  - [Webinars](https://www.ibm.com/think/webinars)

- Reports






  - [Cost of a Data Breach Report 2025](https://www.ibm.com/reports/data-breach)
  - [The 2025 CEO Study](https://www.ibm.com/thought-leadership/institute-business-value/c-suite-study/ceo)
  - [IBM X-Force 2025 Threat Intelligence Index](https://www.ibm.com/thought-leadership/institute-business-value/report/2025-threat-intelligence-index)
  - [Industries in the AI era](https://www.ibm.com/thought-leadership/institute-business-value/report/industries-ai-era)
  - [Orchestrating agentic AI for intelligent business operations](https://www.ibm.com/thought-leadership/institute-business-value/report/agentic-process-automation)
  - [Scaling supply chain resilience: Agentic AI for autonomous operations](https://www.ibm.com/thought-leadership/institute-business-value/report/supply-chain-ai-automation-oracle)
  - [AI in Action report](https://www.ibm.com/think/reports/ai-in-action)
  - [State of Sustainability Readiness Report](https://www.ibm.com/think/reports/sustainability-readiness)
  - [View all IBV reports](https://www.ibm.com/thought-leadership/institute-business-value)

- Podcasts






  - [Overview](https://www.ibm.com/think/podcasts)
  - [AI in Action](https://www.ibm.com/think/podcasts/ai-in-action)
  - [Mixture of Experts](https://www.ibm.com/think/podcasts/mixture-of-experts)
  - [Security Intelligence](https://www.ibm.com/think/podcasts/security-intelligence)
  - [Smart Talks with IBM](https://www.ibm.com/think/podcasts/smart-talks)
  - [Techsplainers](https://www.ibm.com/think/podcasts/techsplainers)
  - [The Coherence Times](https://www.ibm.com/th

... (truncated, 14 KB total)

Resource ID: 674736d5e6082df6 | Stable ID: ODE2NDkxNT