I’ll be honest that it would have been much more difficult to defend Anthropic against the DoW incursion had that incident occurred after this one. This is the company literally telling their customers, “we reserve the right to silently sabotage you.” I’d still have defended them, because the government trying to destroy a firm is still wrong, but man would it have been a harder case to make.

The best part of all these Claude 5 Fable safety measures is I bet the jailbreaking community will still get past them, so the people doing open research in good faith don't get access to the best models but bad actors maybe can.

welp my vision here was probably wrong and indeed there will be an extreme asymmetry of outcomes

to be clear: i think this is obviously in line both with anthropic's stated humanitarian values and choices around safety and also in line with its economic interests. i strongly expect openai will do the same, although they will may delay it until they have no choice.

@deanwball It starts with frontier LLM development and then it's going to evolve to harness development.

How convenient for competing with API customers who are not ultimate enterprise token end-users.

Anthropic is indeed a supply-chain risk... to its private enterprise customers.

Take:

@matt_is_nice @paulmarin90 yes

@tenobrus The Fable release is basically incapable of teaching users anything helpful, from what I'm seeing it blocks interactions from anyone interested in anything to do with ML, bio or chem, even at kindergarten levels.

@deanwball @paulmarin90 Is it really silent if they’re coming out and telling you they might do it though?

@deanwball but they have disclosed in a report - is that so bad?

@ianwsperber I don’t develop frontier models, but I absolutely do ML and LLM research and engineering on a regular basis using coding agents.

@deanwball I wasn't under the impression you developed frontier AI models. Anthropic's intentions here seem entirely legitimate to me. I'm glad if they are preventing distillation or "rogue RSI." If this hinders generic research on current LLM design then I might change my mind.

@deanwball Can't do algebra or analysis for linear mixed models (not ML, but feels overly repressive)

@_ankur7 @deanwball It’s horrible because you’ll never know when it’s doing it. A huge erosion of trust.

@g5p8v8r7s7 @_ankur7 @deanwball Users who don’t like that possibility can simply stop using Claude, correct?

@ianwsperber Doing eg alignment research on frontier models *is* frontier ai development!

Fully agreed! As a private company, Anthropic has the right to design its products and negotiate its contracts with its customers however way it pleases free without the government trying to nuke it.

Likewise, business can evaluate the risks posed to them by each vendor and proceed accordingly.

@matt_is_nice @deanwball Yes. And I don't think we should give them too much credit for this disclosure....not disclosing this behavior could reasonably be interpreted as false advertisement or fraud.

If not in a court of law, then certaintly in the cout of vibes.

@idhe838370042 @deanwball LOL

@deanwball @timhwang Claude Judas.