Nobody Knows Who Built AI Coding Model Ox Alpha or Where the Code Goes

My take: An anonymous AI model called Ox Alpha appeared on OpenRouter on August 20 with no company name, no privacy policy, no safety commitments, and a claimed 80% score on the DeepSWE coding benchmark, above Claude Fable 5 (65%) and GPT-5.6-Sol (52%). Free, for a limited time. The temptation is real.

But the question that matters isn't how high it scores: it's where your code goes. SiliconAngle reported that nobody knows who built this model. Worth noting: independent researcher Ben Davis ran the full DeepSWE benchmark separately and found results much closer to GPT-5.6-Sol mid, meaning the 80% figure may reflect a specific subset, not the full evaluation.

If you use a free model like this without knowing who's behind it, your prompts, your code, and whatever you share go to a server you can't audit. For casual exploration or public-facing code, that might be fine. For professional work or proprietary code, it's worth a second thought.

Do you have a clear policy on which AI tools you'll use with your company's data or your clients' code?