AI-powered cyberattack kits are 'just a matter of time,' warns Google exec

CISOs must prepare for "a really different world" where cybercriminals can reliably automate cyberattacks at scale, according to a senior Googler. Heather Adkins, veep of security engineering at the ad and cloud giant, said it probably won't be for a few years to come, but cybercriminals are already using AI to enhance small …

[Rich 2]

AI is great!!

Damn the critics. This is yet another excellent demonstration of how useful and helpful to society AI…..

Oh. Hang on….

[IGotOut]

Well Google...

... you're partly to blame for this soooo....die?

[Blue Screen of Bleurgh]

AI generated hacker-agent hacks AI datacentre farms

AI datacentre agents google AI for AI help as western world goes dark

Microsoft CEO insists this is just a blip, and an update is on its way and AI is here to stay etc.

[Anonymous Coward]

Malware families are already using LLMs

"You may ask for one prompt to be granted, as celebration on this, the day of my daughter's wedding."

"Eh, why you want to ask me *that*? On this, the day of my daughter's wedding?"

"I answer this prompt for you, you gotta be ready to, you know, reciprocate, one day, maybe sooner, maybe later; capiche?"

[Winkypop]

Fast transition

AI will go from “Meh” to “Mega dangerous” in very quick succession.

[Anonymous Coward]

Re: Fast transition ... Slow Regret !!!

'AI' is already 'Mega dangerous' the responsible 'meatsacks' have just not noticed yet !!!

By time it is noticed it will be too late.

Just like exploding all those 'atomic weapons' in the high atmosphere cannot be undone ... so letting everyone have free-reign to 'play' with 'AI' is going to come back to bite us all.

The need to get 'AI' out into the world at 'any cost' means that 'AI' will be used to impart harm by the 'usual suspects'.

People will look back at the stupidity of the Tech Bros to NOT regulate 'AI' when it could !!!

'AI' has little use for the public at large BUT can be used to concentrate 'efforts' on hacking systems and creating the largest amount of harm possible with limited resources.

In this instance, 'AI' does NOT have to be correct 100% of the time as long as it is correct some of the time and can therefore peform the actions needed some of the time !!!

The analogy is having a large box of 1000 Grenades where 90% are faulty and will not explode if you 'pull the pin'.

This does not matter if you can pull all the pins in 30 seconds, the ones that are working will be enough to cause the damage you want !!!

:)

[amanfromMars 1]

Re: Fast transition ... Slow Regret !!!

The need to get 'AI' out into the world at 'any cost' means that 'AI' will be used to impart harm by the 'usual suspects’. ...... Anonymous Coward

The undoubted fear which is surely gripping and absolutely terrifying all manner of ‘usual suspects’, AC, is AI able to be efficiently used to impart a great deal more than just harm to the aforementioned ‘usual suspects’ by all manner of unusual and anonymous subjects .... considerably smarter autonomous entities.

And the slow dawning of the fact that AI is here to stay and is to reign supreme and play no second fiddle to any lesser operator with a skewed and unpleasant alternative agenda is another virtualisable future reality to get used to and to prosper in rather than try to oppose and deny has any validity and credibility. Carpe Diem.

[Eric 9001]

Prompt a model

To commit playful cleverness against a company? What?

If it is possible to crack a companies computers by shoving the output of a LLM model though an API that does things, then literally anyone with any cracking ability could crack such computers without a LLM - I don't see what difference LLM's make.

While that might allow a script kiddie, that can't even run scripts, to take a bunch of stuff down, the one at fault for that would clearly be the maintainer of such hopelessly insecure systems.

[Anonymous Coward]

All of this could have been less threatening

If companies actually wanted (a decade ago) to hire enough entry level soc analysts to train them to become good talents in the future.

But no, companies prefer to shout that people don't want to work anymore, blame Gen-Z for not wanting to suffer having to work as HelpDesk representatives or any shitty bullshit jobs (which are hell on earth for autistic individuals such as myself.)

Or hire recruitment specialists that believe that displaying bad social skills. (in interviews for any technical careers that require minimal/no work relations with tech illiterate customers.

Is grounds for not being a good fit for the role (technical roles don't require soft-skills. Why the fuck do companies insist on this stupid requirement yet complain that autistic people such as myself don't enter the workforce and waste tax-payers money.)

These corpos brought this upon themselves. And I don't give a shit anymore. I tried enough times to "climb the corporate ladder". It's bullshit.

Why hackers exist? Because HR DOESN'T WANT TO HIRE PEOPLE WITH "BAD SOFT-SKILLS" AND "LOW AMOUNT OF FRIENDS ON LINKEDIN".

SO FUCK ALL OF THESE COMPANIES. BECAUSE OF THEM I WORK A SHITTY HELPDESK JOB BECAUSE I'M DONE BEING UNEMPLOYED AND CHASING NEW CERTS AND PROJECTS IN CYBERSECURITY. THEY ALL FUCKING DESERVE THE DATA BREACHES AT THIS POINT.

I had to fake being neurotypical to get this shitty job btw.

[amanfromMars 1]

Re: All of this could have been less threatening if human leadership was not so unintelligent

It appears that a Carpe Diem time has come for you to rejoice, Autistic Anonymous Coward. ....... https://forums.theregister.com/forum/all/2026/01/23/ai_cyberattack_google_security/#c_5218187 ....... albeit not in praise of any way anyone was expecting.

And yes, there is absolutely no doubt it can certainly very easily be more than just extremely threatening and cataclysmically dangerous ..... existential life on Earth threatening ........ for such is an available choice option in the many arsenals of novel and noblel goodies and monumentally destructive quantum leap weaponry at ITs disposal.