Most "AI tools" are someone else's model with a logo on it — and whoever owns that model, and whatever laws reach it, reach your data too. Vannus grades all 373 on model ownership and data sovereignty, quoting the vendor's own words wherever they publish them. Paste your stack and see how many sit within reach of the U.S. CLOUD Act. No paid placements.
Our scores aren't for sale. We prove it in code Corrections & right of reply.
Vannus is an AI tool resilience rating service. It grades 373 AI tools A+ through F on whether they will still be worth paying for — specifically whether a tool builds its own foundation model or is a thin wrapper reselling someone else's (GPT, Claude, Gemini). Most "AI tools" are the latter; when the upstream model changes, the wrapper's value can evaporate.
What makes Vannus different from a directory: every grade that turns on which model a vendor runs traces to the vendor's own documentation. 253 of the 373 tools carry a citation to the vendor's published page. A further 97 were established to run no foundation model at all, so there is nothing to cite — that is an answer, not a gap. For the remaining 23 the product uses AI and the vendor does not name the provider, and the entry says exactly that. Every tool in the catalog now sits in one of those three states. We are not aware of another directory that publishes the source behind each grade — if you find one, tell us and we will say so here. Scoring is also structurally walled off from affiliate revenue — enforced by a test in the build, not promised in marketing.
The methodology weighs nine dimensions of trust: data sovereignty, allied infrastructure, training privacy, conditional privacy, compliance, operational resilience, exit portability, real-world utility, and a caution flag for material risks. The free catalog and the "grade my stack" tool cost nothing. A self-serve Exposure Report on your own stack is a one-time $299. Stack Watch, a monitoring subscription that alerts you when a vendor changes what it runs, is $149/month and cancellable at any time. A founder-led Concierge audit of a full AI stack is $7,500.
373 AI tools graded on resilience. Paste your stack or search by task.
253 of 373 grades trace to the vendor's own documentation, and a further 97 products run no foundation model at all — 18 of those cited to the vendor, the rest established from what the product is. Where a vendor does not disclose what model it runs, we say so instead of guessing — and never mark the vendor down for it.
Founder-led audit of your full AI stack. 14-day deliverable. Pre-screened intake.