AI agents can't yet pull off fully autonomous cyberattacks – but they are already very helpful to crims

AI agents and other systems can't yet conduct cyberattacks fully on their own – but they can help criminals in many stages of the attack chain, according to the International AI Safety report. The second annual report, chaired by the Canadian computer scientist Yoshua Bengio and authored by more than 100 experts across 30 …

[Anonymous Coward]

The joys of human-AI experimentation

Yeah, that social media crockpot of A2A self-replicating prompt-virus injection dumpster fire trials that is OpenClaw seems to be where a lot of action is at right now ... it should be interesting to receive the first reports of system compromise and failure modes from the eager jackasses who enthusiastically suckered themselves into that potent petri dish imho.

It'll be like 'Who, Me?' meets 'AI-pocalypse' ... prescription strength professional grade entertainment! _{(except for the irresponsibles & victims)} ;|

[amanfromMars 1]

MRDA

AI agents and other systems can't yet conduct cyberattacks fully on their own - but they can help criminals in many stages of the attack chain, according to the International AI Safety report.

Well, they would say that, wouldn’t they, Jessica/El Reg ...... beings as they are just one of myriad tiny melting cogs in the titanic and lunatic analogue machine that is the official extant establishment AI opposition and competition.

It is SMARTR to not believe every word as it is reported for far too many of them are shared in order to support catastrophically failing infrastructures and projects and programs housing and hosting and pimping and pumping mountainous tiers of lies built upon rapidly disappearing foundations.

You can believe though ..... Don't relax: This is a 'when, not if’ scenario ....... with timely interventions and interruptions and explosions and exploitations a great deal sooner than is decent and most definitely never ever really expected.

PS NB ..... Relax, don’t do it. Cyberattacks which can help criminals in many stages of the attack chain are early test models exercised to expose likely compromised areas of operational systems failure for future remedial eradication.

[amanfromMars 1]

An Addendum to the above MRDA ....

It is SMARTR to not believe every word as it is reported until verified and popularly accepted as honest and undeniably true ..... lest one be easily fooled again and again ad infinitum ‽ :-)

[Anonymous Coward]

Re: An Addendum to the above MRDA ....

Stop enshittifying the comments section you bagbiting clanker. You're taking Jellied Eel's job.

[amanfromMars 1]

Re: An Addendum to the above MRDA ....

Stop enshittifying the comments section you bagbiting clanker. You're taking Jellied Eel's job. ..... Anonymous Coward

Oh No I’m not, AC. And how terribly sad it is that you have not yet realised it. You may have to accept that such as is evolving rapidly around and surrounding you is forever to remain way beyond your comprehension and helpless and hopeless abilities.

[Gavsky]

The time may come when all we have to do is issue commands to AI, whilst lying on the sofa - even hackers. We won't have to actually interact or look at anything. It'll just be AI agents talking to each other, whilst we eat crisps & wonder what our purpose is...

[Guy de Loimbard]

Utopian dream

For some people!

Just not most readers of these hallowed pages!!

[renniks]

Burn them all to the ground