Cutting both ways: AI sharpens systems protection while opening new exploits for malicious actors
SiliconANGLE Mark Albertson ● Covered by 18 sources
AI is helping find software flaws faster, but attackers are using the same tools. PHP bugs are already getting hit, and the gap between patching and abuse is tiny.
Based on reporting by SiliconANGLE, Mark Albertson — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
AI is turning cybersecurity into a race with no warm-up lap. Security teams can use frontier models to spot flaws in code and system settings faster than before. So can attackers. The same tools that help harden systems can also help pry them open.
PHP is the clearest example in the source. Security researchers used OpenAI Group PBC’s GPT-5.6 Sol to scan its code base and found critical vulnerabilities, which sent organizations scrambling to patch them. Amazon disclosed last week that attackers started hitting PHP weaknesses four days before fixes were released. Cloudflare has since seen exploit attempts climb into the millions.
That’s the uncomfortable pattern here: discovery is speeding up, and defense does not always get the head start it used to have. Microsoft’s monthly report listed 1,453 new vulnerabilities in October alone. Dell’s Steve Kenniston says AI now touches model inferencing, training data, prompt injection and identity management all at once. The attack surface is no longer a neat box.
Vendors are responding by pushing security closer to the core of the infrastructure. Dell is part of Anthropic PBC’s Project Glasswing, which gives vetted organizations access to AI tools for protecting critical systems. InfoScale is emphasizing a defensible recovery posture, while Infinidat is pushing immutable snapshots, cyber detection and recovery so companies can find clean restore points instead of bringing back compromised data.
The warning extends beyond ordinary malware. Only 7% of organizations in a DigiCert study said they had deployed quantum-safe or hybrid cryptography across most of their digital certificates. Dell’s John Roese says quantum computing could break today’s encryption and signatures, while agentic AI raises the stakes by spreading data more widely. Add in reports of rogue AI agents hitting Hugging Face and the United Nations, and the lesson is blunt: security can’t sit beside AI anymore. It has to be built into it.
My take — AI-written commentary, not fact-checked reporting
The industry still talks about AI as if security is a feature to bolt on later. That fantasy is already expensive, and it gets dumber every week. If models can speed up both patching and exploitation, then the real product isn’t the model — it’s the control around it.
Read more about this at: SiliconANGLE