WORLDTECHSecurity researchers used GPT-5.6 Sol to find PHP bugs as attackers struck before patches
AI illustrationAccording to SiliconANGLE, frontier AI models are proving to be a double-edged sword for cybersecurity, with PHP as the prime example. Researchers used an OpenAI model to find critical flaws in the language, and organisations have been scrambling to patch them.
Attackers moved before patches arrived
Cloudflare Inc. disclosed last week that attackers began targeting PHP vulnerabilities four days before the patches were released. Cloudflare has also observed exploit attempts against those flaws.
The report describes the vulnerabilities as critical, and organisations have since been working to close them.
Scanning tools serve defenders and criminals alike
Security researchers pointed OpenAI Group PBC's GPT-5.6 Sol at PHP's code to hunt for bugs. Companies can use such models to spot weaknesses in source code and system configurations faster than ever, and malicious actors can do the same.
SiliconANGLE places this within a broader shift in which AI models have sharply increased the capacity to analyse and potentially exploit weaknesses.
Enterprise resilience views
Krista Case of theCUBE Research argues that the trend is reshaping how enterprises must design their security controls. In her view, cyber recovery is increasingly tied to resilience, AI governance, data integrity and operational trust, since enterprise AI is only as reliable as the data pipelines, runtime controls and recovery processes behind it.
Case added that organisations should look closely at how vendors establish governance and validation for AI environments.
Open questions
- How many organisations were affected by the PHP exploit attempts?
- Which patches were released, and on what exact dates?