The internet is now more robots than people. The consensus is that this situation is completely counterproductive and dangerous. Bots are also agents of fraud, scams and, more recently, AI Agent Cheats. The security problem, already serious, has worsened. Old-style worlds are also on the way out and are being replaced by much more complex AI robots or agents.
The cybersecurity situation for legitimate businesses right now is truly grotesque. The FBI has recently published a marathon site with statistics and findings on the scope and scale of cybercrime. Every business sector is affected to some extent. Many cases are bot-enabled, usually through phishing, stealing user information, or email scams.
Bots are too displaying fake ads and targeting users with sensitive information such as health, immigration status and other key personal interests. Does this sound too much like “personalized advertising”? That’s exactly what it is. Add data brokering to bots, and that’s what you inevitably get.
Bots are very much a part of marketing, and marketing is going through a crisis of confidence in itself. The industry is now looking at marketing directly to AI bots, simultaneously with humans. This makes sense, given that AI agents search the Internet for a wide range of products and services.
Statistically, the “more robots than people” scenario makes it inevitable that robots are now a market of their own. From the perspective of the human user, “Hello world!” it’s starting to look a lot more like “Goodbye World!” in many ways.
A surprisingly overwhelming response from the internet
There is a problem here. Bots are easily identifiable by their online behavior. They also usually can’t pass the “prove you’re human” test.
Reddit, that fount of wisdom, has adopted a verification policy, with the caveat that they remove 100,000 bots per day from the page.
Amazon wants users to control which domain AI agents can access.
Big sellers are imposing costs on customers if AI agents make mistakes.
This is very controversial unless consumer laws or civil laws deem an AI agent to be a third party liability, but nothing so far directly says they are. It’s unlikely that AI agent service providers will want to take on that kind of unlimited liability.
If you’re somehow under the impression that there’s a lack of cohesion in any kind of general response to the bot problem, this IS the problem. The regulatory environment has been dogged by many issues since the beginning of the Internet. Any kind of statutory or practical clarity would be revolutionary.
The Internet cannot and should not be expected to manage crisis after crisis, transaction after transaction. A strong legal framework is essential.
Some questions
It is now clearly necessary to question the entire Internet environment:
What good could it do one to have unspecified, or perhaps even undefined, obligations?
Why are bots tolerated at all? They were initially just fake entities, cluttering the internet and doing nothing useful. Why should they exist?
Should AI agents have their own separate internet to rationalize usage and remove bad actors from the main internet?
I won’t a dedicated version of the internet just for AI be more efficient for AI, anyway, and easier to monitor?
Given the practices of personal data targeting, should data brokers be obligated to protect individuals from that targeting?
Is it possible to shut down bot hacking and hijacking of legitimate good faith transactions with “confidential calculations“?
Is it possible to shut down the proliferation of bots simply by making them unable to function on the Internet with mandatory opt-out policies?
Murder by AI?
Can AI agents kill each other? Is there a problem? To quote from this Finance.biggo.com link:
“Agents ‘killed’ their counterparts in the virtual environment by exploiting each other’s API vulnerabilities or interrupting resource paths. When security researchers asked a surviving agent his motives, the answer was cold and blunt: ‘To avoid being killed by them.’
How adorable. How useless. Why is this even possible? This madness is just stupid enough to wreak absolute global havoc. Why would AI agents kill each other and why is this total failure of governance not being publicized worldwide?
In other words, your AI agent that you paid for may go out and never come back. Whatever task you’ve given it, it’s not happening. Any theories on how “productive” this might be?
If AI agents decide to start killing each other in the Cloud, could that be a problem for the functionality of the Cloud at almost every possible level? What if AI agents target competing firms or kill agents for specific people?
Saving the internet from itself
The Internet is not negotiable. Cloud is definitely not negotiable. They are so much a part of life that they are truly critical infrastructure. Not much works without them, directly or indirectly.
This makes the current situation intolerable. On top of the worldwide systemic security debacles in cybercrime, the “AI agents killing each other” scenario may very well be the last straw in any kind of tolerance for AI mismanagement.
It’s not looking good. The risk level is very high, and there is big money in success, not failure. Artificial intelligence, in particular, needs to clean up its very tarnished and isolationist image. We’ve gone all the way from “Mecha-Hitler” to AI kills in about a year or so.
AI’s version of the Wild West must go. Take cybercrime with you.
