imagine telling your customers there's a small chance you'll randomly decide they're using your product wrong and you won't tell them but will secretly silently sabotage their work
humans& co-founder Eric Zelikman criticizes tech companies for silently sabotaging user workloads under the guise of product moderation
Zelikman is a former xAI contributor and Stanford researcher.
Many users condemned AI firms for secretly nerfing models and gaslighting customers without notice, calling it ecosystem sabotage and a sign that users are no longer the real priority.
Most Activity
haha yeah that's crazy. anyway time to use Grok to get the news
imagine telling your customers there's a small chance you'll randomly decide they're using your product wrong and you won't tell them but will secretly silently sabotage their work
I would like to +1 that this is a very bad policy. Respond with a refusal and deal with the fall out but invisible NERFing is super uncool.
Who on Earth would do something like this
imagine telling your customers there's a small chance you'll randomly decide they're using your product wrong and you won't tell them but will secretly silently sabotage their work

@_oleh nobody would perform such an own-goal, right?

@ericzelikman oh really? you've tried to reverse engineer our model training? fine, now let's see if you can reverse engineer if our model is gaslighting you or not

@ericzelikman Huh that’d be crazy. Surely no one would want to give their competition a free win like that

@ericzelikman @yacineMTB actually this has been happening since Claude Code was released… they are just being more transparent this time around

@ericzelikman @mattparlmer i hate to break it to you but we’re not their customer anymore, the USG is

@ericzelikman

@ericzelikman and you get to pay them for the experience

@Suhail Anthropic are assholes.
Fable is useless, denying all requests cutting usage in half and cutting off users in 2 weeks.
A woke, retarded, douchebag company.

@Suhail Respond with a refusal or not.... just overall worrisome trend

@TheNewPolicy I do not think they are assholes but they will not get everything right all the time so it's important to steer them to the degree we can which I think is best done in a constructive way.

@ericzelikman .@_sholtodouglas @bcherny could someone over there please address this? i have a ton of empathy for safeguards, but this feels kinda wild..

@ericzelikman @jeremyphoward So fucked up. @FTC

@ericzelikman They've been doing it for years. They're just admitting it now. But they've been sabotaging AI researchers for years. It was noticable to me a longgg time ago in my own work.

@bayeslord The kind of people who run popcopy

@ericzelikman @yacineMTB Makes perfect sense why they pissed off the Pentagon.

@ericzelikman Invisible degradation is the worst of both worlds: you keep the risk while deleting the feedback loop. If you need a boundary, make it explicit so users can adapt their workflow.

@ericzelikman It's easy if you try