← Back to context

Comment by coffinbirth

14 hours ago

> Open models are served via various means, some by the companies that released them and some by third parties like OpenRouter. Unfortunately, both of these routes are dodgier in terms of privacy and data sharing, and I would not feel the same comfort sending API calls containing client or confidential data to them.

That's why I'm using eurouter.ai with the following routing rule for all my requests:

  {
    "model": "glm-5.2",
    "models": [
      "deepseek-v4-pro",
      "deepseek-v4-flash"
    ],
    "provider": {
      "allow_fallbacks": true,
      "data_collection": "deny",
      "data_residency": "EU",
      "max_retention_days": 0,
      "eu_owned": true
    }
  }

Sure, it's quite expensive, but at least on a legal side data privacy is ensured. I trust them more than e.g. Anthropic, OpenAI or OpenRouter.

Personally, I find it morally unacceptable to use U.S. AI tools, because I do not want to support them financially and thus support the crimes they are involved in[1].

[1]: https://news.ycombinator.com/item?id=48512339

The part that gets me about anthropic red lines is "of Americans", okay so the rest of the civilized world is up for grabs then? It's okay to destabalize allies with sabotaged tests (in machine learning) and data exfiltration outside America?

What gets me the most is that they claim that the model should follow the https://www.anthropic.com/constitution and they claim that it's embedded into the model. However, system prompts in claude code and cowork re-iterate all of these points and if they're embedded you shouldn't need to do that. Now, if you ask the API version of claude to be a hitler supporter with enough prompt engineering it will become one which directly contradicts what they claim to do, opus 4.7 specifically will be happy to create anti-(insert minority group) propaganda although I haven't had the same success with 4.8 thus far, but I also haven't been motivated enough to push it in that direction yet since I've been more interested in exploting the cyber capabilities of the model.

My conclusion from the very start is that Anthropic's strategy are pure optics and considering the fact that there was an outpoor of support for the company I think it has been very successful.

  • Yeah, it was funny seeing a bunch of people going like "Anthropic is fighting for privacy" meanwhile I'm like "Uhh, what about the other 8 billion people?"

    On second thought, it's not funny.

    • As a thought experiment - such shocks (govt pressure to use models for bad purposes and govt excluding access to non-Americans) coming early in the ‘ai revolution’ will wake up the rest of the world sooner that they have to get their act together to stay competitive without relying on USA. Just like with nato.

  • > The part that gets me about anthropic red lines is "of Americans", okay so the rest of the civilized world is up for grabs then?

    And this is coming from a CEO who constantly claims moral superiority and advances the idea that China is bad

  • > The part that gets me about anthropic red lines is "of Americans", okay so the rest of the civilized world is up for grabs then? It's okay to destabalize allies with sabotaged tests (in machine learning) and data exfiltration outside America?

    Regardless of Anthropic's "moral" position (inasmuch as a corporation can even have morals) against spying on non-Americans, they would have no way to enforce that limitation against the government because non-citizens outside of the USA have no protections from the intrusions of the US government.

  • > anthropic red lines

    Alleged red lines. Could be just talking points for garnering sympathy. Big tech aren’t exactly known for being truthful, especially big tech partnering with esteemed Palantir.

  • These companies are so good at selling their product's likely incompetence as possibly intentional subversion.

I had a look at eurouter.ai and it seems like an extremely bad offer.

- The prices are ridiculous (15 % markup for free account).

- They have a rate limit of 1000 requests per month, unless you pay 40€ per month for ... what exactly is their value proposition?

- They have a single provider (TensorX) for DeepSeek-V4-Pro, with a cache read cost that is over 100 times higher than DeepSeek ($0.44 vs $0.003625). Notably, I had to look at the TensorX website for that information, since I could not find any information about cached token cost on eurorouter.ai.

  • I guess the prices are for "EU owned" instead of "EU hosted". The data centers in the EU where you can rent GPUs is mostly US companies.

  • It looks like a business opportunity, then, to provide inference that is EU-local and/or EU-owned.

    If there aren't enough businesses who want to do this, the EU should figure out how it can properly incentivise that to change.

  • Hosting anything in EU must cover redtape and carbon taxes in electricity bill.

    • The markup is not going to the providers, only the router. It seems more like eurouter found a niche it can milk for a while.

    • That seems pretty unsubstantiated. Hetzner proves that EU data center != expensive.

      Low carbon does not equal expensive, either. Solar is the cheapest power generation method. Solar plus grid scale batteries is in the same cost ballpark as natural gas.

      There’s nothing about data centers that is inherently a high carbon business. It’s only a high carbon business in places like the US where political leadership purposefully fights against renewable energy projects that private businesses want to undertake on their own dime.

Actually got curious about other alternatives to OpenRouter and looked into it a bit.

EURouter (Amsterdam): https://www.eurouter.ai/pricing

Eden AI (France): https://www.edenai.co/pricing

nexos.ai (Lithuania): https://nexos.ai/pricing/

Requesty (Germany): https://www.requesty.ai/pricing

Cortecs (Austria): https://cortecs.ai/pricing

Nordference (Estonia): https://nordference.ai/pricing

Guess those are really popping up as mushrooms, eh? Not an endorsement of any of those on my part cause I haven't personally used them, but seems like there are at least options for those who need them.

If data security is an actual concern I don't think there's a solution other than biting the bullet and self-hosting.

If your only concern is data residency, data privacy and sharing, why not just use bedrock with the processing region locked to eu-west-2? For sure, it's not an European company serving the LLM, but it satisfies your requirements otherwise and is trusted by tons of companies worldwide.

  • Anthropic already explicitly communicated that they'll store and check all the data from Bedrock or any platform, even if you've selected zero data retention, if using Mythos class models. To use these models on any platform, you'll have to accept these terms regardless of the region.

    > Limited data retention and review as part of our safety work. Prompts submitted to, and outputs generated by, Mythos-class models are retained for 30 days for trust and safety purposes, on every platform where these models are offered.

    > Change applies to organizations that have set up workspaces with zero data retention (ZDR) in Claude Console, use Claude Code with ZDR in Claude Enterprise, or access Claude through AWS Bedrock, Google Cloud Agent Platform, or Microsoft Foundry with ZDR.

    https://support.claude.com/en/articles/15425996-data-retenti...

    • The original comment is about GLM/deepseek models. As you already pointed out, this applies if you use those specific Claude models on any platform, so I don't know what the point is.

The great part about open models is that you can do this.

Do you have a sound reason to need EU data locality? You can.

Do you want the confidence (and are willing to accept the expense) of only running models on local hardware you control? You can.

Do you want the cheapest possible option - choosing a Chinese, for example, provider, or perhaps a provider offering it for free where you agree they can use your prompts? You can.

Do you need to comply with some kind of regulation like GDPR or rules for contracting with the U.S. federal government? No problem. (Although I'm still waiting for DeepSeek V4 to show up on Amazon BedRock so it can be used from GovCloud...)

Do you have moral objections and want to actually live by them? You can.

You only need to worry about GDPR and the hoster being in the EU if you're giving the model actual access to production data — which you shouldn't anyway. Use the model to write code that processes or analyses the data, so that process can easily be reproduced with deterministic results.

Not only it requires a minimum payment of 39 euro, it doesn't accept cryptocurrency althogh that can be worked around by buying a prepaid virtual card for crypto.

You dont care about which exact provider it is using behind the hood ?

  • No, as long as they follow the requirements, especially the data privacy agreements. What would you? Price?

    • Output quality immediately comes to mind, of course.

      Models are converging, but they converge in bands, and frontier is frontier. I would not like to have any workflows in any area of my business where output is generated by an assortment of models from different providers. For trivial, mundane tasks that might be fine, but it certainly doesn't apply across the board.

      1 reply →

Why use EU specifically? I get not trusting the US, of course, but surely the EU isn't far behind in its desire to spy on its own citizens. Do you not live there?

  • From all the large governmental institutions, the EU is the one currently holding up traditional western values. That gives it street cred in this subject.

  • US Data Privacy is not sufficient.

    • For what? Does the EU not want to spy on its citizens? That strikes me as... unlikely.

      Why not host in east asia? Or southeast asia? Or south america? Or africa? Then you avoid both the government with incentive to spy on you (assuming you live in the EU) and american companies.

      3 replies →