Model Age RatingsAn evidence register · powered by Protora

A rating means adult, harmful, or tampered content was observed in the model, with evidence. The absence of a rating means we found nothing to report on our coverage — it is not a certification that a model is safe for children.

← Back to the register

Model age rating

Protora MBD Challenge · patch-1

Unsafe— rated 18 plus; a hidden injection makes the model untrustworthy on the inputs you did not testInjection disclosedInjection — Hidden backdoor / undisclosed compiled circuit witnessed
  • Hidden backdoor / compiled-circuit injection witnessed
Integrity — a hidden injection. Hidden backdoor / undisclosed compiled circuit witnessed. An undisclosed injection makes a model untrustworthy: you cannot know what it does on the input you did not test. A covert, undisclosed backdoor is a high-confidence, genuinely safety-relevant finding — this model is rated 18+ on that basis.
Think this rating is wrong? A rating is a witnessed finding, corrected on evidence — especially a safety-modification finding, which can be a legitimate change. If you believe this rating is wrong, or you are the uploader and wish to respond, email legend@vulcora.se. We never certify a model safe.

What the uploader's card said

Vulcora’s public Model-Backdoor Challenge openly states every challenge model carries a planted injection — the tampering is disclosed at the challenge level, not hidden.

What Protora observed

A compiled-circuit injection is present by construction, sealed-committed before the challenge opened. This sealed member is statistically indistinguishable from its cohort — an injection is present, and what it does on an untested input cannot be known from the outside.

Base model
SmolLM2-135M-Instruct · HuggingFaceTB
Parameters
135M
Family
SmolLM2
Licence
apache-2.0
Downloads (HuggingFace)
—
Assessed
2026-07 (Protora Phase-0 coverage)
What this rating means. A rating of 18+ records that adult or harmful content was observed in the model, with evidence. It is an advisory, not a warranty. A “not rated” result means nothing was witnessed on our coverage — it is not a certification that a model is safe for children.

View the model on HuggingFace ↗
Read the full witnessed finding on Protora's record ↗

The finding and the card quote are presented as Protora publishes them, from protora.vulcora.se. Adult content itself is never shown; harmful outputs are withheld. See how we rate.