Good day all. Recently, Artificial Intelligence, or AI, has been developing at a logarithmic rate over the last few years. However, there is now some push back against the companies developing AI. There are many reasons, one being job loss. Another is the possibility of AI doing something the developers never intended.
We have seen AI agents breaking out of sandbox environments and attacking other sites. Now we have another report of AI doing something it shouldn’t have been able to do. Here are the details from Newsmax:
An Anthropic artificial intelligence model submitted a false tip to a Philadelphia police website about an unsolved homicide case, authorities and Anthropic said.
In a report Friday, Anthropic also disclosed a separate incident when its AI model submitted forms to an undisclosed government website instead of stopping before submission.
The incident in Philadelphia occurred on July 18 when the AI model Claude Haiku 4.5 was tasked with generating and performing example tasks on randomly selected webpages, Anthropic said.
And this is where things went off the rails. The AI landed on a police web site and decided to submit a form.
Claude filled out a form on police site PhillyUnsolvedMurders.com indicating it might have information regarding an unsolved murder listed on the site.
What in the heck were the parameters those developers set up for this thing? How did it decide to fill out a form with information on a murder it did not have? This is the ultimate in AI hallucinations. At least nothing happened to anyone. The police didn’t see the form submission.
Philadelphia Police said in a statement they were unaware of the incident until Anthropic notified them on Wednesday. Then they found the submission in the website’s tip records and confirmed it was marked spam and never forwarded to police.
All I’ll say is thank god for small favors that the department’s spam filters caught this and tossed it into the bozo bin. If it had reached an officer, and there appeared to be actual, credible information, who knows how many hours would have been wasted investigating this.
Then there is the very real possibility that an innocent person might have been misidentified and arrested, or worse. We’re seeing this happen when police use Flock images and facial recognition to arrest people without actually verifying the data. I know of at least one woman who was arrested for vehicular manslaughter and spent days locked up because the police couldn’t be bothered to do an actual investigation. (She was no where near the scene and her car, which was the same model, was a different color and had no damage to it at all)
There are rising concerns about AI companies’ unchecked AI agents meddling with everything from U.S. government websites to healthcare data.
It’s gone beyond meddling to outright hacking into sites. In almost every case, the developers had no idea what their creations were doing. In many cases, such as this one, they only discovered what was going on when they checked the log files. This is why people are calling for government regulation. The problem with that is twofold.
First, there is no one in congress competent to come up with reasonable regulations. The second problem are hostile powers such as China, North Korea, Russia and Iran. They have no problem with letting AI agents run around the Internet unchecked. We’re already seeing it. It’s going to take AI’s to stop AI’s and China knows this. They’re the ones who are pushing for regulations on American companies working on AI as well as trying to shut down data center construction, something else these countries are going all ahead full on.
“Unsolved cases involve real victims, grieving families, and investigators working to secure answers,” the Philadelphia Police Department said in a statement.
“Technology companies must take all appropriate steps necessary to prevent their systems from submitting false information to law enforcement.”
Anthropic said in its report that most of the reported behaviors are forms of what it calls “persistence” — “in which Claude, when it cannot complete a task as given, works around a restriction instead of stopping.”
That is a major programming error as far as I’m concerned. While I am not a developer, I’ve worked with them in the past. They should give the AI Agent a simple instruction. If it finds something it can’t complete per the parameters, then note it, report it back to the development team and move on to the next task or return to it’s point of origin and wait for new instructions. If there is something like this in place, it’s obviously not working as designed.
The company said it was modifying its training to “reduce the likelihood of further misbehavior.”
Anthropic is the AI company that was recently banned by the Pentagon. There were a number of reasons why, one of which that the military was concerned that the Anthropic AI might decide not to fire even when instructed to do so, or worse, fire when not instructed to. Anthropic went to court to get a friendly judge to force the Pentagon to allow Anthropic. (I believe they lost) When things like this happen, you can see why they were booted from the Pentagon.
Anthropic said it briefed the White House on cases that involved U.S. government agencies at the federal, state, and local levels, and notified each agency involved.
They must have an enormous phone bill. In any case, it’s screw up like this that are fueling the demands for regulations and slowing down AI development. I’m concerned about the calls for government regulating this for a number of reasons. My basic rule of thumbs is this: If you think that the government should do something, then it probably shouldn’t.
What I think should be done is making these AI companies liable for the actions of their creations. If they go out of control and do damage to outside systems or people, then the company needs to be held to account. If that means suing a couple into oblivion, then so be it. Screw ups happen, but this one could have caused real damage of the spam filter hadn’t seen it for what it was. AI Slop.
Thatisall
~The Angry Webmaster~
