Today · post, October 26

AVERI reviewed 800 Amazon safety prompts; Amazon chose them

AVERI reviewed 800 prompts from Amazon's internal AI safety test set and found some refusals may hinge on surface wording; Amazon is changing four things

Our AI analysis, unconfirmed: An outside review of Amazon's safety tests is a new format, but Amazon still controls the sample and the fixes Check back when when AVERI publishes its full report or a second company opens a similar review

Concept illustration: on the left of a desk stands a tall stack of papers wrapped by a band lettered "AMAZON" with an orange curved line under the letters, and a label reading "SAFETY QUESTIONS" points to the stack by a line; in the centre a person with dark hair in a bun, a black turtleneck and a badge on a lanyard holds a sheet of paper in both hands, with a label reading "AVERI" pointing to them by a line. On the right of the desk are two shorter stacks of papers, the rightmost one topped by an upright red rectangular block.

Why it matters: An outside group examined Amazon's safety tests and published the weaknesses. Amazon chose the 800 prompts, described the fixes and had approval rights over publication, so the review says little about how safe Nova 2 is.

AVERI, a nonprofit that pushes for third-party AI audits, published its review of Amazon's internal safety test set on October 26 (AVERI, 2026-10-26). AVERI reviewed the quality of the tests, not the models, using 800 prompts Amazon supplied. Some attack prompts test only safety filters, and some refusals may be driven by surface cues rather than safety reasoning. Amazon responded on October 20 with four changes: removing redundant prompts and adding harder ones, grading by difficulty, expanding languages and quarterly reviews (Amazon Science, 2026-10-20).

Not proven yet: Amazon supplied the sample, and AVERI didn't verify how it was drawn How we checked →

Judgment update: Our reading: this review examined the test set, not the model, and the reviewed company held approval rights. That gives it more weight than a company's self-assessment but less than a third party testing the model directly. Amazon alone described both the sample and the fixes. So the review can improve how the test set is written, but it adds no new data on whether Nova 2 is safe. AVERI's full report comes next. The other open question is whether a second company opens its internal test set to outside review.

Source and checks

This item ran in the brief of Oct 27, 2026, alongside 12 other items. How we checked it, the quote checks and what was still missing that day are on that issue's How we checked page.

Cite this item

SecondSource (2026). "AVERI reviewed 800 Amazon safety prompts; Amazon chose them." https://secondsource.io/en/news/2026-10-27-3/. Accessed access date.