Skip to content

Articles

Private AI, backed by evidence.

One question per article, answered with dated facts, named organisations and primary documents. Every source carries the day it was read.

38 articles

  • AI infrastructure

    What an fp8 key-value cache changes, calculated

    The key-value cache is the memory that grows with every person using the server, and its precision is a setting, not a property of the hardware. Calculated on a 96 GB card: 185 conversations at fp8 against 104 at bfloat16, with the same weights and model.

    3 min read

  • Data, privacy and GDPR

    DeepSeek writes down where data goes. A regulator wrote the exception.

    The privacy policy says the data is processed and stored in China. Korea’s regulator found that user input went to a cloud company in Beijing. Berlin named the GDPR article. The Czech regulator excluded locally deployed open models from its own warning.

    5 min read

  • AI infrastructure

    A local AI benchmark published two numbers. Only one travelled.

    A vendor measured 54 tokens per second at 32K context and 96 seconds to the first token on the same machine. The effective rate across the whole response is 8.9 tokens per second cold and 53.2 warm. What a buyer should ask for instead.

    3 min read

  • Private AI

    What is private AI, and when does an organisation need it?

    Private AI is the same kind of model with a different answer to one question: who controls the environment where the data is processed. What that answer changes, what it leaves untouched, and the five questions that settle it.

    3 min read

  • Private AI

    Private AI or cloud AI? Read the provider's own terms first.

    The comparison is usually framed as secure against insecure, and that framing loses arguments. The honest comparison reads the contract: what the provider promises, what the deployment type changes, who may read a flagged prompt, and what the buyer still owns.

    4 min read

  • Data, privacy and GDPR

    Shadow AI: the tool arrives before the policy does

    Shadow AI is what happens when the demand for a tool arrives before the approved version of it. The measured scale, why a ban produces the worst version of the outcome, and what an approved path has to contain.

    4 min read

  • Private AI

    What an on-premise AI appliance is and what it saves you from building

    An AI appliance is the decision not to become an AI infrastructure company. The eleven components a local deployment needs, the one measured example of how much a wrong choice costs, and what an appliance gives up in exchange for removing the assembly.

    4 min read

  • Private AI

    Seven questions to answer before you decide where AI runs

    Start with the workload, not the hardware. What the system will read, whether the data has to leave, how steady the usage is, whether the internet is needed, how strong the model has to be, who operates it, and what happens when it is unavailable.

    4 min read