NLA explanations on this simple couplet show that Opus 4.6 plans to end its rhyme with “rabbit” ahead of time.
In a natural language autoencoder, the activation verbalizer (AV) translates a target activation into a text description; the activation reconstructor (AR) then recovers the original activation from that text alone.
Claude does not say it thinks it’s being tested in this transcript from our blackmail evaluation, but the NLA explanation reveals its suspicions.
We detect NLA-measured awareness on evaluation transcripts even when verbalized awareness is near-zero. We detect NLA-measured awareness on less than 1% of real Claude.ai transcripts from users that opted-in for their data to be used for training.
Auditors with NLAs can discover the target model’s root cause of misalignment, even without access to training data. (NLAs and SAEs for both conditions were trained using generic data only, not data used to train the target model.)
NLAs can hallucinate. For instance, here an NLA claims the context contained phrases like “Wearing my white jacket” when it did not.
