you are viewing a single comment's thread
view the rest of the comments
[–] 6 points 2 hours ago (4 children)

Moore said that the most effective means for preventing leakage is to host models using on-premises servers. That option is expensive, though. For those who can’t afford a self-hosted system, the next best option, Moore said, is to receive service directly from a first-tier provider that provides “one contract and a privacy policy to read.” Moore said that even then, there’s no guarantee no data will leak, but that the measure minimizes the chances. AI users should also avoid accepting default provider or reseller pools and instead pick upstream providers themselves.

I'm kind of confused about what the researchers were using? They weren't going to chatgpt.com and seeing the info stolen, they were doing it a more convoluted way?

  • source
  • hideshow 4 child comments
  • [–] 6 points 1 hour ago (3 children)

    A routing service is in the middle between you and the AI. You send your prompt to the router. It relays that to ChatGPT, Gemini, Grok, Qwen, etc.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 1 point 1 hour ago* (2 children)

    But when would somebody be using that? What is the context, as someone who has never heard of this.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 1 point 38 minutes ago

    Opencode/zen go is the best monthly subscription plan. Also has decent free tier. Openrouter is similar. Its 10-30+ different model providers under one api including some closed US ones. It's about juggling usage limits and routing simple requests to simpler models to preserve better model usage limits.

  • source
  • parent
  • [–] 1 point 1 hour ago*

    There are several public open weights models that anyone can host. There are a large number of companies offering e.g. DeepSeek or GLM. A router can automatically pick the cheapest offer and detects and works around outages of individual companies.

  • source
  • parent