What it means in practice
Imagine that instead of buying a finished car from a dealership (ChatGPT from OpenAI, say), you are given the plans and the engine for free. You assemble the car yourself in your own garage. It takes more technical skill, but the car is entirely yours and nobody tracks where you drive it. The best-known open-source model today is Metaβs Llama.
Why companies need to know this
- Data protection (GDPR). If you process sensitive client data, you must not send it to public servers in the US. An open-source model runs on your hardware and the data never leaves the company.
- No licence fees. You pay no monthly charges for API calls, although you do pay for your own servers and running them.
- Full control. You can adapt the model to your business exactly, and nobody can switch it off on you overnight.
An example from practice
A sales company processes sensitive management reports containing internal financial data. Sending those to commercial ChatGPT was unacceptable from a security standpoint. The answer was deploying an open-source model on their own company cloud. The AI helps them mine the data just as well, without the security risk.
In our company AI workshops we teach teams to work safely with company data so you do not give away your competitive advantage.