XOR Information

XOR is a defensive cyber lab. We sell reinforcement learning environments and benchmark data to frontier labs and coding-tool companies. Their agents train against verifiable exploit targets and learn to find and fix vulnerabilities at production scale. What we do for AI enterprises: 1. Benchmark AI agents against real problems before you buy or deploy them 2. Builtin the audit trail your board and regulators are going to ask for anyway 3. Help you effectively deploy AI Agents which are worth paying for We work with large enterprises who are deploying AI agents - but help smaller teams, too. We are backed by NVIDIA, AWS, Google and CISPA Helmholtz center - the #1 ranked Cybersecurity Institution (CS ranking). Companies are buying AI security tools faster than anyone can check if they work. These tools write code, patch vulnerabilities, make decisions. And most of the time, the only proof they work is the vendor's own marketing. That is a terrible way to make a six-figure purchasing decision. We built an independent benchmark. XOR runs AI agents against hundreds of realworld vulnerabilities from production software. OpenSSL, FFmpeg, curl. Not toy problems. Not simulations. The same bugs that cause real breaches. We tested 15 of them. The best fixed 61%. The worst fixed 26%. Nobody knew which was which until we published the data. Now they do. Our testing methodology follows an internet standard from the IETF. Same standards body that created HTTPS. Our co-founder is co-chair of two IETF working groups for the past decade. He wrote the rules before we built the product. If you are spending money on AI security and want to know what you are getting, we should talk. Talk to the team: https://xor.tech/contact