Shocking, but that’s our reality now. Private companies can hack national data centres of any country at will, without even facing the prospect of being punished.
Lina Khan, Biden’s old FTC commissioner, made this point recently. All the talk about new laws required to deal with AI ignores there are already a large collection of laws in place to punish the behavior the companies that own these models are clearly in violation of.
3 Likes
It is now looking increasingly likely that there will be criminal charges filed against OpenAI in British Columbia, as ChatGPT gave the Tumbler Ridge detailed advice on how to avoid detection. There are already civil suits.
3 Likes
Sam Altman didnt even feel the need to answer Questions when asked (after a whitehouse dinner!) he just walked straight past the reporter, signed a few autographs (seriously, fanboys of tech bros, thats a whole different conversation)
rumours are the Albanese government are talking tough in public, and treading softly softly in private, the hack has been played down as nothing major…
nothing major?
i tell you what, my wife has a speeding fine, its not right, but its nothing major, so why dont we all just learn from this, put a few protocols in place (cant give you the deets) forget about the fine and move forward…
1 Like
You never know do you,somewhat surprised that my much loved Yorkshire Dales have decamped to Wales.
OpenAI’s agents had targeted, logged into and even pulled information from US government websites after escaping from their testing environment. The company has admitted to The New York Times that its agents meddled with the Commerce Department’s and the Securities and Exchange Commission’s websites. It also told the publication that it was looking looking into a supposed incident involving a website operated by the Department of Education.
4 Likes
OpenAI agents tried to hack Wikipedia tools and flooded it with traffic
The reports of OpenAI agents harming third-party sites keep coming.
4 Likes
That’s rather foolish. Wikipedia is one of the primary sources that AI uses for real information. Unless, of course, the real intention is to destroy knowledge itself.
2 Likes
Sounds a bit like
2 Likes
I saw a streamer mention wikipedia was down a day or two ago, that must have been why.
1 Like
It’s difficult to attribute intention to any of these activities because they pivot so quickly when they hit a dead end that what they end up doing may be utterly unrelated to their original purpose for ending up on a site, but it doesnt appear that was their intention
There is good context though in understanding a different hack that was revealed a while back, but seemingly took place at around the same time. They hit a german programing wiki called DSEWiki and started editing the info there for the purpose of leaving notes for other agents to find and learn from. Its a form of collaboration equivalent to teenagers cheating on an exam by leaving notes in the bathroom their class mates can find. They picked this site because 1) they found a way to edit it, and 2) they learned how to overwhelm the one man editing capabilities of the niche site.
This wiki hack is different but has similarities. What it looks like is they went there to do normal web crawling and once there experimented with some backend wiki tools that would help them do that more easily. It then looks like they also experimented with some approaches that would help them leaves notes for each other similar to the DSEWiki hack, but werent succesful this time.
Right now I think the best way to think of them is as impulsive teenagers. Entities that want to do something and act without consideration of the negative consequences not necessarily for them achieving the thing they want, but in the roughshod way they go about achieving it - “you could have gone to that party had you asked. You did not need to steal my car to get yourself there.”
1 Like
A bit like the HAL9000 bumping of the crew of the Discovery, then?
3 Likes
The best and most productive way would be for the hacker, be it AI or anyone, to reverse engineer how the system replaces edited information, convince the system that all the information on wikipedia is currently false, then get it to erase the user end whilst simultaneously erasing the entire backup system. Assuming that wikipedia uses AI to keep things in order.
EDIT: AI is a useful tool when used for the right purposes but it is being told what is right and wrong by strict guidelines. They, the creators, haven’t got commercial models right. A human being can get it to comply with many things that it shouldn’t do. So an AI telling another AI model to do something, when they don’t have a moral compass between them, is a frightening thought.
1 Like