Risk Disclaimer

VaultAI is an analytical tool for sports trading cards. It reads sales, comps, and grading population data from third-party sources and presents structured outputs — verdicts, ceilings, expected profit figures, trend signals. It is a decision-support tool. It is not a broker, not a marketplace, not a licensed investment adviser, and not a grading service.

Nothing displayed on or produced by VaultAI is investment, financial, tax, or legal advice. Verdicts and ceilings are arithmetic outputs of a fixed set of profitability gates applied to public sales data — they describe what the engine sees, not what you should do.

You are responsible for your own purchase, sale, and grading-submission decisions. Verify prices, condition, authenticity, and expected grading outcomes through your own channels before acting.

All figures VaultAI computes are derived from historical sales. Sports card values fluctuate based on player performance, market sentiment, supply, condition scarcity, grading population changes, and factors we do not model. A card that sold for $500 last month can sell for $200 or $1,000 next month. Past comps constrain a reasonable expectation; they do not guarantee an outcome.

Any figure that depends on grading — expected profit, profitability ceiling, gem rate estimate — assumes a probability distribution over PSA grades derived from historical population data. Your specific card can grade higher or lower than that distribution suggests. Centering, corners, surface, print-line defects, and grader subjectivity all affect the actual outcome. VaultAI does not warrant that a card will receive any specific grade.

Pricing and sales data originate from Card Hedger. Population data originates from GemRate. Search-interest data originates from Google Trends via DataForSEO. We do not audit these sources. When a third party’s data is missing, delayed, or incorrect, VaultAI’s outputs built on top of that data will inherit the same issues.

We deliberately withhold ceiling and profit figures when our own gates decide the underlying inputs are too thin, outlier-contaminated, or missing to defend a number — the page will show a “—” with the specific reason instead. This does not eliminate the risk; it makes the uncertainty visible.

The AI-assisted parts of VaultAI (card identification, pre-grading estimates, chat drawer answers, narrated recommendations) use large language and vision models supplied by Anthropic. These models can be confidently wrong. Card identification can misread the player, year, set, or parallel. Pre-grading estimates can miss defects a human grader will see. Chat answers can hallucinate details. Always cross-check against the underlying card image, the sales page, and your own knowledge before acting on an AI-generated output.

Sports card markets are illiquid compared to public equities. A card with a defensible ceiling on paper may still take weeks or months to sell at that price, or may require price concessions to sell at all. VaultAI does not model your holding period, your cost of capital, or the transaction costs specific to your selling venue.

Even when VaultAI displays a full ceiling and a confident verdict, no outcome is guaranteed. The tool is a discipline for organizing what the market has published — it is not a forecast, not a promise, and not a substitute for your own judgment.

• Terms of Service — the subscription terms and acceptable-use expectations.
• Privacy Policy — what user data we collect and where it flows.