Let me get this one thing clear right off the bat, insurance for agents doesn't mean insurance for insurance agents. It means insurance for AI agents, and I think that this is the new lucrative model that the valley has been able to come up with where I see a few of these companies trying to capture this space of giving insurance and risk assessment for when AI agents fuck up.
I think it's a very interesting business model and business problem to solve, because historically, I think that insurances have had a very static rate and in the era of these AI models, I think it is getting Even more. imperative than I can ever stress on it to have dynamic premium rates. What I mean by that is, since an agent can sort of improve its performance. Either the agent owner does it, or the agent does it on its own, doesn't matter. The ability of the agent to carry out multiple tasks increases. I won't say increases, but for a lack of words, I think it gets more efficient and it makes errors lesser times. It hallucinates less, it drifts less. So, the risk associated with that sort of an agent going to fuck up gets lesser, which means that companies can inherently charge a lesser premium for that sort of ensuring that sort of an agent. But here's where it gets interesting. And AJ that is highly risky can still be insured by these companies. Just at a higher premium.
Now, I don't know where these companies draw the line at something they are going to ensure or not because there are gonna be agents that have implications in the real world as well as agents that don't. Let's take for example, a trading agent. A trading agent is basically going to take the capital that is provided with and traded across a variety of financial instruments and try to get the best arbitrage and the best deal for the person who created that agent naturally. Now, it all really depends on the person who created the agent, to make the agent a high risk, high reward machine, a lower, low risk, low reward machine. And the owner entirely lies on the insurance companies, to sort of benchmark these agents. understand the intent behind these agents, see if it's the right. fit for them. And then ensure that. In a very broad sense, then these companies become more of an investment company where they are actively evaluating certain use cases and certain efficiency levels of agents to see whether it's a right fit for them to ensure. They wouldn't want an agent that fucks up too often to get insured and they wouldn't want an agent that fucks up too less for them to be able to charge a very down market premium for that agent.
The problem gets even more interesting when you think about it from an ethical point of view, right, and this, I'm mainly talking about agents that do have real implications in the real world. Which are going to be ages that help with maybe medicine research, right, or agents that help with, uh, dosing patients for clinical trials. I don't know if there are any yet, but I think there will be, and the insurance market is going to be used for them because in the end, the question then doesn't become who made these agents, but rather who is going to take responsibility for when these agents mess up. Naturally, it would be the companies or orchestrators of the agent, but since and they can command disrespect, since the orchestrators or companies have actively paid premiums to ensure that the agent. When it does fuck up, extracts are covered, and doesn't follow them back to the owner is quite a controversial future and a very questionable one if you come to think about it because the consequences of that agent messing up mean that there was a problem with training, but also mean that there is probably a bigger problem. For the people who got exposed to those clinical trials or that new cancer treatment.
And I think to set up these companies is gonna be a huge burn machine, like existing investors are gonna have to put in hundreds of millions, if not billions, to even get these things off the ground because you really have to provide extremely lucrative premiums to sort of get people actively involved in your services and their only bet is, or their core thesis lies in the fact that when you expose people do less accountability, they are naturally not gonna want to go to a future or to the past where they had more accountability in terms of the agent's outcomes. And so once you give them a taste of the sweet candy, they're not gonna want to have it taken back from them, which is why then you can start delving into higher premiums. And when, You can think of this like a network effect. But when about 60% of the agents out there that are doing some meaningful work either for a person, an organisation, or just humanity in general, are gonna be insured against their hallucinations. Then the 40% that is not insured is going to want to get insured too because they get put in a bigger spotlight for messing up and they get to have to pay larger fines or larger consequences as compared to their competitors, if I may. And that would in turn lead to a much larger loss in profit or revenue, or output value in general, that they're gonna want to run the numbers and really see if the insurance is gonna be worth it. And you never know how many times your agent messes up, it can be once a week, once a month, once a year, once in five or maybe even never, right, but. The question is who's willing to take that risk?
Most of you would think what are normal insurance companies gonna do you know the ones that gave insurance to humans? I think it's gonna also be a very huge opportunity for them, not because some of them are going to venture into this space of providing insurance for AI agents. But for, like I talked about, for the humans getting affected by these AI agent actions. I think a large part of the human population who is currently on an insurance plan is going to want insurance against consequences caused by I agents. I think that's gonna be a really new hot premium plan that is gonna just fly off the shelves when insurance companies put it in place. Given that there is a really good adoption for insurance for AI agents. Because, I mean, if you're ever in a jam that you think is caused by an AI agent, you can just sort of put a claim on that insurance, and even, you know, you can get paid for it. Now there's always gonna be, like, 2 sides of the same point where people are actively going to try to fall prey to these AI agents, apparently naively, to sort of get the claim on their insurance premiums against consequences caused by agents. But for humans, and I mean, there's always gonna be the people who are actually affected, and then these legacy insurance companies, like they always do, are going to want to, in some way, not give heed to the latter class and void all those insurance claims, like they've been doing for years because apparently that's how they make money.
Its going to be very interesting to see how this entire new vertical of policy making takes shape and how much widespread adoption it has, especially within the tech world.