Topic starter
18/08/2026 2:52 pm
TL;DR: Pre-Mythos models still enable end-to-end exploit development, according to Intruder, after its engineer used a token-driven workflow to find a SQL injection flaw affecting 300,000-plus sites, and says the practical constraint is operator skill rather than model availability. The lesson for defenders is that AI-assisted attack capability is already present, so exposure reduction and patch speed matter more than debating model bans.
Editorial analysis by NHI Mgmt Group, based on content published by Intruder: “The Vulnerabulletin Issue #8”.
Key questions
Q: How do security teams reduce risk from AI-assisted exploit research?
A: By assuming research timelines are shorter and by tightening the time between disclosure, validation, and mitigation.
Q: Why does blocking one AI model not remove offensive capability?
A: Because the capability lives in the workflow, not in one named model.
Q: What are the warning signs that AI-assisted scanning is being used against you?
A: Look for rapid-fire requests across many endpoints, repeated payload variants, and quick transitions from reconnaissance to exploit-like requests.
Practitioner guidance
- Tighten exposure management for internet-facing systems Inventory externally reachable assets continuously, then prioritise the apps and services that can be reached without internal controls.
- Accelerate remediation for high-probability flaws Set patch SLAs by exploitability and reachability, not by CVE count alone.
- Harden detection for automated exploit development Watch for bursty scanning, repeated payload variants, and fast follow-on proof-of-concept activity that suggests tool-assisted research rather than casual probing.
Bottom line: AI hacking tools still matter because skilled operators can chain discovery and exploitation even when flagship models are restricted.
Explore further
View Full Forum → | NHI Foundation Course → | Our Services → | Read the full analysis →
18/08/2026 3:36 pm
AI-assisted offence is now an operator-skill problem, not a model-availability problem. The article’s core point is that restricting one public model does not remove the underlying exploit workflow when capable operators can reproduce the same outcome with other tools. That shifts the defensive question from whether attackers can access a named model to whether they can still chain research, validation, and exploitation fast enough to matter. Practitioners should treat capability transfer as the real risk, not platform access alone.
A few things that frame the scale:
- AI-related credential leaks surged 81.5% year-over-year in 2025, with the surrounding AI infrastructure leaking 5x faster than core LLM providers, according to the State of Secrets Sprawl 2026.
A question worth separating out:
Q: Should organisations prioritise patch speed over debating AI tool bans?
A: Yes, when the main threat is public-facing vulnerability exploitation. Tool bans may matter for policy, but they do little against operators who can use other models or workflows. Patch speed, asset inventory, and exposure reduction directly change whether a found flaw becomes a real intrusion opportunity.
👉 Read our full editorial: AI hacking tools still matter because skilled operators do more with them