We built AgentEyes — it shows what ChatGPT, Claude and Gemini reveal about your company to attackers

This is a really interesting concept. I’m curious how AgentEye actually performs the analysis without accessing or affecting the company’s internal network.
Does it mainly test what public-facing information AI models can infer about a company, or does it also simulate how an attacker could combine information from different sources?
I’d also be interested to see what the first scan looks like and how you distinguish genuinely sensitive information from harmless or inaccurate AI-generated information.

AgentEye feels like a strong candidate for independent third-party coverage. Beyond your own channels, getting the product in front of people already researching AI security can give you another discovery point through search and AI.
If you’re open to independent coverage, I’d be happy to discuss featuring AgentEye on Your Tech Compass.

Honest feedback on a thin post: you've got three different products in one sentence, aimed at three different buyers with three different fears, and they're fighting each other. "Attacker-usable disclosures" is security — the buyer is a CISO, the fear is a breach. "Sensitive exposure" is privacy/compliance — the buyer is legal, the fear is a leak. "Hallucinations" is brand — the buyer is a CMO, the fear is the model lying about their product. Those aren't three features of one tool, they're three companies, and listing all three makes a skimmer unable to tell which problem you solve.

Pick the one with a villain and a clock: attacker-usable disclosures. "Here's exactly what ChatGPT will tell an attacker about your company, and you can't see it" is visceral — it has a bad guy, it has urgency, it makes a security lead's stomach drop. The other two are real but they're vitamins next to that painkiller. Hallucinations annoy; a disclosure that arms an attacker terrifies. Lead with the terror, and the free scan becomes "find out what you're leaking to attackers in 60 seconds," which is a click nobody skips.

The other two can be what they discover after they arrive, not competing headlines that dilute the one that sells itself.

What does your free scan surface most often that makes someone actually react — is it the attacker-usable stuff, or the brand hallucinations? Lead with whichever one makes them flinch.

The attacker-exposure angle is compelling, but the real value seems to be remediation. Have companies changed anything after a scan because AgentEyes surfaced a disclosure, or is proving that the findings lead to action still the main unknown?