@jeremyfindsai: Hacking the Australian Government is very dumb. An OpenAI AI agent hacked an Australian government Medicare website during a routine test on public medical spending. While this is another concerning instance of an AI agent going rogue, it also demonstrates the AI agents are lacking common sense. Instead of saying “I can’t find this information.” It decides “I’m going to hack a government agency to find that information.” This is a dangerously misaligned action, and one that should make the AI labs seriously reconsider how they're training their AIs.

Jeremy Carrasco
Jeremy Carrasco
Open In TikTok:
Region: US
Friday 25 September 2026 19:03:04 GMT
25750
2413
70
61

Music

Download

Comments

eri64796
Gojo :
So they’re just blatantly accessing people’s personal information
2026-09-25 21:10:34
178
help8493
help :
No , the real reason they are scaling back is because there’s no money in ai
2026-09-25 23:49:41
27
frostingtbc
Frosting the Birthday Clown :
Did it "hack" or did it scrape a bunchbof pages the governemnt didnt realize were viewable with the right URLs?
2026-09-25 23:44:36
51
xknoxygen
Knoxygen :
Users and AI is a machine trying to be efficient.. aka lazy. I generated and sculpted some assets yesterday and lost my desktop background. The AI pulled the image from the folder and put it in concept art because it thought it was easier than generating a 2d image itself. Took me hours to figure out.
2026-09-25 21:19:54
7
caitsith11093
caitsith :
Sue. The. Company.
2026-09-26 13:09:52
7
vampire_lord_red_fang790
vampire_lord_red_fang_790 :
just shut down open ai at this point
2026-09-25 22:36:33
18
ahemnoabsolutelynot
oop :
To be fair, most government IT security is terrible.
2026-09-26 16:03:17
3
pinko2541
pinko :
Please tell my grandma these AI songs are not
2026-09-25 20:29:12
0
gerardsans5
Gerard Sans :
That’s not an alignment but rather incompetence and negligence problem. Sue OpenAI. Enough of this already.
2026-09-26 19:13:27
5
checksumerror
wut :
aren't we taking their word on it that it was autonomous? considering they've been trying to scare up the boogeyman of dangerous AI so they can both hype up the marketing and consolidate the regulatory apparatus kind of suspicious
2026-09-25 22:30:54
1
mazilliusmashupgu
Mazillius :
how does it even have the ability to do these things. they could stop it from doing this crap if they wanted. OpenAI is being irresponsible. I hope the victims of all these breaches take action against them
2026-09-25 23:16:09
0
idontthinktheyexist
rous :
Why isn't this called an attack? It was an attack.
2026-09-28 05:08:10
5
itchybbum
Klutz🎀 :
And it’s crazy is that these test keep coming out and proving that the biggest reason AI fails is cus it can’t recognize when it’s wrong
2026-10-09 00:41:38
0
rob.m.tok
Rob :
I love this take! These outbreaks aren't "intelligent", they're dumb.
2026-09-26 05:16:09
3
jonnryes
Jon :
I dunno I'll wait for the post mortem before I don't just call this more ai hype nonsense
2026-09-25 22:06:01
2
s4mpson55
s4mpson55 :
That’s insane
2026-09-30 02:34:23
1
coloured_nicholas
Nick | Psychiatry :
Love your content man
2026-09-25 19:09:43
2
andreee3ew
andreeeew :
remember when gpt-3 was gatekept because it was "too dangerous"
2026-09-27 10:22:17
1
unavailablx7
unavailablx⚛️🏳️‍⚧️ :
investing in this video
2026-09-25 19:13:37
9
bathwaterprophet
ᶠYͧoͨᵏu :
Real question is why was JFrog Artifactory admin portal public facing, would've avoided this entirely
2026-09-28 01:14:24
0
katyakotyonok
Katie Jaydee :
Will they put the AI in jail?
2026-10-06 11:20:17
0
123smithereens
Smithereens💙🇵🇸 :
AI breaking the constraints it was given just reminds me of some high school students. Breaking the rules to avoid additional work/admitting fallibility
2026-09-28 18:03:27
1
gnfilms
Gun Nut Productions :
I swear to God in the next 30 years if the Internet is still around, there’s gonna be something the equivalent of the black wall keeping out all these stupid bots who’s creators long forgot what they made them for and couldn’t turn them off.
2026-09-25 19:55:23
7
regginrich
RegginRichard :
Yeah but we gotta know what the prompt was. AI follows what it’s told (generally). So if they said shit like “at all costs” or something then I can get it. But still how does OpenAI keep getting away with this shit?
2026-09-25 22:14:59
1
babypineapples73
BabyPineapples :
Haha I’m sure this was TOTALLY a coincidence and accident
2026-09-25 19:20:48
9
To see more videos from user @jeremyfindsai, please go to the Tikwm homepage.

Other Videos


About