Experts at Civo Navigate tell companies to control their AI connectors

At Civo Navigate in London, a panel spoke about open source and AI. Amanda Brock of OpenUK was the chair of the panel. The experts told that companies must have alternatives between models. They also told that AI code contributions are too many for the maintainers of open source projects. Liz Rice of Isovalent tells that attackers frequently use bugs before a person knows of them.

Models and connectors

Matt Barker of Jetstack tells that frontier labs can want to control all the connections to your enterprise. Companies want alternatives. Barker tells companies to use the open Model Context Protocol, a standard that connects AI tools to data and tools. Anthropic made it and gave it to the industry.

Security and upkeep

Rice tells that a patch is not sufficient when attackers use a bug before the patch is available. Rice wants runtime checks, which stops attackers while software operates. Rice tells that AI agents cannot keep projects without a person.

What is not known

The panel gave no numbers or test results. Kelsey Hightower tells that some experts want a ban on Chinese open models in US organizations. Hightower tells that other nations can think the same of the US.

Insight: When models are interchangeable, the connector is the prize

Dell tells that 80% of AI tokens at this time come from open models, but it gave no data. Barker tells that frontier labs can want to control all the connections to your enterprise. If models are easy to change, the control point reads as a move to the connector. The one that controls it keeps the customer. Monitor who controls the Model Context Protocol.