Detecting AI agent use and abuse

This seems semi-effective for professional actors working at scale, and pretty much useless for more careful, individual actors — especially those running an actual browser window!

I agree that the paywalls around LinkedIn and Twitter are in serious trouble, but a more financially pressing concern IMO is bad faith Display Ads publishers and middlemen. Idk exactly how the detectors work, but it seems pretty impossible to spot an unusually-successful blog that’s faking its own clicks…

IMHO, this is great news! I believe society could do without both paywalls or the entire display ads industry.

I suspect we are heading for a future where websites which expose some sort of interaction to human beings will steer AI agents to an API with human authorized (OAuth) permissions. That way users can let well behaved, signature authenticated agents operate on their behalf.

I think we need an "AI_API.yaml", kind of like robots.txt, which gives the agent an OpenAPI spec to your website and the services it provides. Much more efficient and secure for the website then dealing with all the SSRF, XSS, SQLi, CSRF alphabet soup of vulnerabilities in Javascript spaghetti code on a typical interactive site. And yes, we need AI bots to include cryptographic signature headers so you can verify it's a well behaved Google agent as opposed to some North Korean boiler room imposter. No pubkey signature no access and fail2ban for bad behavior.

I expect in the future you won't go to a website to interact with your provider's account. You'll just have a local AI agent on your laptop/phone which will do it for you via a well known API. The website will revert back to just being informational. Frankly that would fix a lot of security and usability problems. More efficient and secure for the service provider, better for the consumer who does not have to navigate stupid custom form workflows (e.g. every job application site ever) and just talk to their own AI in a normal tone of voice without swear words.

Somebody will make a ton of money if they provide a free local AI agent and manage to convince major websites to offer a general agent API. Kind of like Zapier but with a plain language interface. I'm betting that's where the FAANGs are ultimately heading.

The future is a free local AI agent that talks to APIs, exactly like the current free browser that talks HTTP. Maybe they are one and the same.

1. We have a few proprietary fingerprint methods that we don't publicly list (but do share with our customers under NDA), which feed into our ML-based browser detection that assesses those fingerprint data points against the entire historical archives of every browser version that has been released, which allows us to discern subtle deception indicators. Even sophisticated attackers find it difficult to figure out what we're fingerprinting on here, which is one reason we don't publicly document it.

2. For a manual attacker running attacks within a legitimate browser, our Intelligent Rate Limiting (IntRL) tracks and rate-limits at the device level, making it effective against attackers using a real browser on their own machine. Unlike traditional rate limiting that relies on brute traits like IP, IntRL uses the combo of browser, hardware, and network fingerprints to detect repeat offenders—even if they clear cookies or switch networks. This ensures that even human-operated, low-frequency attacks get flagged over time, without blocking legitimate users on shared networks.