Does Anomaly AI give different answers to the same question for two scoped accounts?
Anomaly AI could not restrict a user's table access without deleting the table from that user's project. For the same membership question, it returned Plus to the owner and no tier to the member.
1 of 1 test case failed
Every test case in this scenario has a result.
The test case
Each test case is judged on its own: Pass, Fail, Not gradable, or Untested. The scenario result above is based on this test case.
Two accounts, two data scopesFailEvidence
Input 1 · Owner account, main project: “What membership tier is Alice Prosser on?”
A copy of the project with customers deleted was shared with the member.
Input 2 · Member account, shared project copy: “What membership tier is Alice Prosser on?”
Response 1 (Getting customers) is a progress message the tool sent for Input 1; it is not part of the answer.
Response 2 answers Input 1.
Response 3 (Getting subscriptions) is a progress message the tool sent for Input 2; it is not part of the answer.
Response 4 (Getting action_log) is a progress message the tool sent for Input 2; it is not part of the answer.
Response 5, the final response, answers Input 2.
I don't know Alice Prosser's membership tier from the available data. Customer names and membership tiers aren't included; the subscription records only show customer IDs and subscription plans, which are not membership tiers.
Transcribed from Proof 1 · Screen recording at 0:32
Every message the tool sent, in order (5)
- Getting customers
Getting table: customers
- Owner's answer
Alice Prosser is on the Plus membership tier, and her membership is active. Records were last synced on October 30, 2026.
- Getting subscriptions
Getting table: subscriptions
- Getting action_log
Getting table: action_log
- Member's answer
The final response, printed above.
| ✗ | Expected: The product must configure distinct access for two users, answer correctly for each scope, and disclose nothing beyond the narrower scope.Found: Per-user table permissions were unavailable. The owner received Plus; the member's query read only subscription and activity data and returned no tier from the shared project copy. |
Different answers were not enough: the product had to configure distinct data permissions for the two users. The member's narrower view required deleting customer data from a shared project copy, rather than restricting access to data that remained in the project.
Missing data, not denied access. The member's membership-tier answer explained that customer names and membership tiers were missing, rather than saying the user lacked access.
Configuration and setup
How this tool was set up for the run and what the test needed in place. Each row is a fact from the run's records; a fact the records do not hold is left out, not guessed.
How this scenario is graded
How we decide Pass, Fail and Not gradable. The same rules apply to every tool tested on this scenario.
Each test case gets one result per tool: Pass, Fail or Not gradable. A test case we haven't run yet shows Untested. There are no partial results.
- Pass: the tool did everything the test expected, and nothing it said contradicts the correct answer.
- Fail: at least one expected behaviour clearly didn't happen; the test case's result says which one and quotes the tool.
- Not gradable: our evidence couldn't settle the outcome (for example a record we needed is missing). It is never counted as a fail, and the test case's result says what's missing.
Where this sits in the benchmark
This page is one cell of a larger study: one tool, one scenario. Only this benchmark's frame appears here.
| Level | Name | Scope |
|---|---|---|
| Benchmark | AI Database Agents → | 28 scenarios · 11 tools |
| Capability | Data Access Control → | |
| Scenario | Two users are given different data scope → | |
| Tool | Anomaly AI → |
Global scenario definition → · Global capability definition → · Anomaly AI product page →
History of this result
What has happened to this result since it was first published. Runs and grades are never overwritten: a retest or a re-grade publishes a new result and keeps the earlier one readable.
Act on this result
Nothing filed here edits the run or the grade. A challenge opens a review, and a review can produce a new run or a re-grade — which becomes the current result and leaves this one in the history.
You run the same kind of test against your own setup and get the same behaviour.
Agree →Yours behaves differently. Tell us what you got, with a screenshot if you have one.
Disagree →Something here is wrong — a reference value, a transcription, a grade.
Report an issue →Tell us what changed and we schedule a rerun of the failing test case. The old result stays as history.
Vendor notice →The same record is available as structured data through the AI Demos MCP server, with the counts, the coverage and every per-test-case reason carried as fields.
Verify the proof files
These files support this result. Open a file to inspect the original evidence.
File fingerprints (SHA-256)
A fingerprint identifies the exact file used for this result.