More on the research point - things are hard enough in external evaluation world already

Anthropic’s move might sound reasonable if you consider their actions as a company chasing superintelligence.

But consider that their customers are spending billions of dollars on their services! That is precisely what has led to their recent surge in ARR, popularity, and fund raising success.

So customers’ surprise and anger is warranted when they then sandbag in evals *without even informing them* about the degraded capabilities.

it seems unethical that Fable would secretly do a shitty job instead of just refusing the request.

If you take customer's money and accept the job, you should do the job right. This is like the baker, instead of refusing to bake a cake for a gay wedding, takes the job, but intentionally bakes a shitty cake

@MTSlive Silent sandbagging is arguably more dangerous than visible refusals—it creates a false sense of capability while obscuring the model’s true failure modes. Do we have any data on how often users detect this behavior in practice?