CoinSec joins Anthropic's Cyber Verification Program

What the programme is

Anthropic's models refuse most offensive security work by default, and rightly so. The Cyber Verification Program is how vetted security organisations get that work unblocked for legitimate, defensive use: finding vulnerabilities, writing exploits to prove them, and testing systems with the owner's permission. Verification is per organisation, after review by Anthropic.

Why it matters for your code

CoinSec attacks your program the way an adversary would, and keeps only the exploits it can prove. That is dual-use work by definition, and until now it meant working around the same safeguards an attacker runs into. Verification lets us use the most capable models for the whole attack, from reading the program to running the exploit on a fork, without the run stopping halfway.

It is also a check on us. Anthropic reviewed who we are and what we do before saying yes. If you are deciding whether to let an AI attacker near your code, that review is one more reason to.

What does not change

The rules we work under. Exploits run on a fork of the chain and never against mainnet. Findings go to your team. If you build on Solana, join the waitlist.