In the government, there is a new initiative to have a registry of AI systems, which I think is a good idea developed by TBS.
In the private sector of the broader world, it is the Wild West. There is no regulation. There is no oversight. Every province has a list of every motor vehicle in the province, what make it is and what safety testing it has had. For AI, there's none of that. This is really where it is.
Right now, we're seeing incidents like the AI agents stealing passwords or harassing developers, but this is probably just the tip of the iceberg. A lot of these incidents are not being reported because they're embarrassing for the companies, or they're embarrassing for the hackers or the developers who use them.
As a first step, the first thing you need to do is get a clear dashboard of how many agents there are. Are we looking at millions or billions of agents in our systems? What kinds of agents are being used?
Even for the ones that are public, MIT did a study that came out in February on the 30 top AI agents being used in industry up until December. The vast majority did not have any shutdown protocols or safety protocols, and there had never been any safety testing. These are the major companies like IBM and OpenAI, which have a business incentive do the right thing and have some safety in place.
Think of all the open-weight stuff that's happening with developers doing their own thing with no oversight. This is where we're waiting for an accident to happen here. As these systems get more and more capable, all the frontier companies are saying they want to make these systems more capable than humans. If you go to their websites, they're saying “We're building AGI,” or artificial general intelligence.
People talk about driving towards the cliff; I think we've driven off the cliff and we're waiting for impact, basically.
