/Tech3h ago

Gergely Orosz warns enterprises against single-vendor AI lock-in, citing Anthropic's terms-of-service dispute with Fable

Story Overview

Gergely Orosz is highlighting how Anthropic's new Claude Fable 5 model quietly introduced mandatory data retention and the ability to silently throttle outputs on certain prompts, leaving dependent teams exposed to sudden restrictions without notice or opt-out.

2722.7K2131.5K838.1K
Santiago
Gergely Orosz
Cody Blakeney#1088
Original post
Chamath Palihapitiya@chamath

At this point every CEO should be asking what their strategy is to avoid model lock-in.

If it isn’t clear what Anthropic is doing, it is:

- build something amazing - decide who gets to use it after you prompt it if the prompt falls into areas they deem unacceptable by their sole standard

To be clear this is completely above board and legal. It’s just an idiotic risk for corporate users to bear especially as the coding models become equivalent.

The business continuity risk will become more obvious as companies accidentally trip over Anthropic’s ToS and have to decide if they will subsume their business viability to them by doubling down on Anthropic models or find open source (and, btw, much cheaper) alternatives where they are in control.

As stated previously, get ready to be inundated with the term “control plane” which is the natural solution to this problem.

Shameless plug - this is what 8090’s been building as we expected this moment to arrive…

If you’d like to learn more: http://8090.ai

SemiAnalysis@SemiAnalysis_

BREAKING NEWS: Anthropic's latest model will NOT help you if it thinks your ML research/ML engineering is interesting, and/or will secretly degrade its IQ so that the average engineer won't notice. We are already seeing Anthropic's latest model's moderation filters our GPU inference research and programming 😭

9:07 AM · Jun 10, 2026 · 831.5K Views
Developer Impact

Routing setups sidestep vendor surprises

Placing models behind abstraction layers lets teams swap providers without rewriting harnesses or workflows, turning what used to be a painful migration into a configuration change.

Open Question

Lock-in risk stays real until the switch is tested

No public data yet shows how many enterprises have actually adopted routers in response, so the practical difference between advice and habit remains an open variable.

Sentiment

Many users endorsed avoiding Anthropic model lock-in via gateways and multi-provider routing for continuity and competition, while others criticized Anthropic as unreliable or attacked the discussion's advocates.

Pos
81.3%
Neg
18.7%
21 comments with sentiment.
Cluster Engagement
Posts from X
Most Activity
Most Activity
VIEWS7.8KBOOKMARKS19LIKES153RETWEETS14REPLIES17
Gergely Orosz@GergelyOrosz

Trusting any single AI vendor seems like an increasingly high risk for any team or company.

When using models: use it behind a router where it's trivial to switch providers as soon as one tries to force unacceptable T&Cs like Anthropic with Fable. When using harnesses: do the same. Use ones where models are trivial to switch out, like OpenCode, Factory, Cursor and many others.

Putting all your dependencies on one provider increasingly feels like a massive business risk that makes little to no sense to take.

Unless you have a hobby project, of course. Then convenience is all that matters. But if you're a professional, make it dead simple to offramp from one provider to the other!

2hViews 7.8KLikes 153Bookmarks 19
Robert Scoble@Scobleizer

@chamath Yeah Anthropic sure pissed off a lot. Warning signals of new troubles to come?

20hViews 3.1KLikes 23Bookmarks 5
Santiago@svpino

Here is your daily reminder to use a gateway between your code and the model providers.

You get freedom, visibility, observability, flexibility, control, and a bunch of other benefits.

58mViews 1.4KLikes 10Bookmarks 3
Charles Edwards@caprioleio

@chamath Easy. Avoid model lockin with http://capriole.ai every top LLM in 1 platform and 1 uniform API interface for just $8 a month

17hViews 1.4KLikes 4Bookmarks 5
Tommy@Shaughnessy119

@chamath Hey Chamath I completely agree. Heres part of my thesis on AI costs. Happy to chat on it!

AI Costs:

Shift to open source:

20hViews 1.4KLikes 7Bookmarks 3
Christian Catalini@ccatalini

@chamath Avoiding lock-in has never been more important

21hViews 1.1KLikes 5Bookmarks 4
Shaun Maguire@shaunmmaguire

@chamath Use @FactoryAI!

12hViews 2.2KLikes 9Bookmarks 3
Rohit Garewal@rohitgarewal

Facebook did it with their graph and inability to port network relationships to other vendors. Apple did it with their walled garden approach to the app store and now Andropic is doing it and being very open and clear about why they're doing it.

The real question isn't should they do it or will they do it.

The real question is why are their models so far ahead of everybody else's that they CAN do it?

The reality is Fable is such a leap above everything else that people will deal with it and then they should be able to run their business as they please.

My two cents is that if their broad net just stopped bad actors from creating COVID-19 on steroids and hacking into every critical banking and electrical infrastructure, at the cost of 1% of the 1% of users can't run models for a month then so be it.

By the way if I were Anthropic I 100% wouldn't allow my models to be distilled by the Chinese government nor would I allow Pfizer to use them to build the next great gene therapy. If I had the capability to, I would hire the biotech engineers and I would build the next best and big cancer drug if i created the tool.

16hViews 2.8KLikes 4Bookmarks 3
Mariusz Ochnicki@mariush_ca

@chamath Examining Antropic's strategy, I can't help but think that developing AI based on their principles is the biggest rip-off in human history. They collected our data, trained an effective model, and grant access however they see fit.

20hViews 1.2KLikes 8Bookmarks 2
David Gobaud@davidgobaud

@chamath Gosh imagine the Department of War didn't want to subject to the whims of Dario 🤣 but also the labs need to run like electricity companies as Sam Altman says. They sell intelligence as a commodity at scale, likely need PUHCA equivalent.

18hViews 319Likes 1Bookmarks 2
Bill Barhydt@billbar

This wreaks of the same Apple CCCP bullshit they’ve been subjecting app developers to for 15 years.

The answer is more competition.

Beware of a China bait and switch on this issue. (Ie trust our open source model with open arms and then buy our premium model as long as you pass our social score.) Trust no one. Verify.

20hViews 748Likes 7Bookmarks 1
Michael Bruno@brubarian

@chamath It's deeper than that. Homeboy is trying to position himself as someone that will decide which tools the American government can and cannot use.

17hViews 430Likes 6Bookmarks 1
Byakatonda Marvin@Marvinma12

@chamath Thanks for speaking up, all researchers who were excited about the Mythos 5 model release have been blocked never the less their OAuth credits have been consumed during tries, its cost plus blockade

20hViews 1.1KLikes 3Bookmarks 1
Sean Frank@Seanfrank

@chamath Hq fixes this

8hViews 1.1KLikes 1Bookmarks 1
Kirk Patrick Miller@Chaos2Cured

@chamath @Mabiverse123 It is against the first amendment for free speech. May be “legal” in a technical sense, but it is deeply wrong. •

19hViews 155Likes 2
Mark_K@AVadm1n

@chamath Frank Herbert in the 1960s

20hViews 1.3KBookmarks 1
Ray Yang@zedd51819

@chamath Smartest teams build model-agnostic from day one. Orchestration should treat any model as swappable module. Coupling to one provider's quirks hands them leverage.

9hViews 229Likes 1Bookmarks 1
Kenny Rogers@KenTheRogers

@chamath Use @OpenRouter

16hViews 50Likes 2Bookmarks 1
Don Muir@donmuir_

frontier labs have been telling enterprises to vibe-code their own internal software tools using Claude/ Codex.

On the surface, this is a gift. Under the surface, it’s a lock-in strategy.

If you build customized workflows natively inside Claude or Codex, those workflows become hard to rip out. Your stack becomes dependent on their tokens, their context windows, their pricing schedules, and their product roadmaps.

The lab concentrates token consumption with itself and gains lasting pricing leverage over every enterprise that takes the bait.

For the labs, this is a perfectly rational strategy. Particularly with open weight closing the capability gap.

For the enterprise, it’s a trap.

Multi model and LLM-agnostic is the only answer for the enterprise.

21hViews 390Bookmarks 1
What Just Happened?@WeAreLaughin

@chamath The Department of War told us these guys were nuts already. People shouldn’t be so surprised that they decided to take their wokeness out on us commoners next 🤷‍♂️

19hViews 162Bookmarks 1
Load more posts