AI agents abound, unbound by rules or safety disclosures

AI agents are becoming more common and more capable, without consensus or standards on how they should behave, say academic researchers. So says MIT’s Computer Science & Artificial Intelligence Laboratory (CSAIL), which analyzed 30 AI agents for its 2025 AI Agent Index, which assesses machine learning models that can take …

[Groo The Wanderer - A Canuck]

I kind of pity the people who are trying to throw everything at AI.

Learning how to do something is what keeps the job fun and interesting.

You're going to be bored 'bot supervisors at best...

I pity you.

[Ken G]

what keeps the job fun and interesting

and some people don't find their fun and interests outside work and just want the money

[Anonymous Coward]

All this wool over my eyes and not a Llama in sight

A bit surprised to see no Llama in this report, except for one mention of PurpleLlama in HubSpot Breeze Agents' 'model protection layer'. Guess we'll have to wait for Avocado, Mango, and Wango Tango, to see what sort of agentic guacamole disaster Meta's kitchen is cooking up ... (not that I'm looking forward to it or anything ;).

[Doctor Syntax]

Safety disclosures. Perhaps this is the answer.

Require all public facing responses - searches, web-site customer disservice bots, emails, images, etc to carry the equivalence of a public health warning

"This material has been created by AI and may be partially or entirely untrue and cannot be safely relied on.".

That should be possible under consumer protection law for anything addressed to the general public and maybe also the likes of Copilot on Windows Home editions. If it were encountered often enough in those circumstances the general idea might seep through to users within business.

[Like a badger]

Perhaps it isn't the answer because people routinely ignore caveats, good advice, warnings.

Imagine a door with a sign saying "Face eating leopard behind this door! Do NOT open". What proportion of humanity will read, understand, and act appropriately? How many people will be suckered today by a romance scam*, or an investment scam** despite warnings since forever?

* About 30, judging by public reports, probably 2 or 3 times that when you include those too embarrassed to report they've been had.

** About 70, plus those who choose not to report.

[Anonymous Coward]

Try painting "why pay to keep this door here? let's feed the cats instead!" on the side of a big bus.

[Doctor Syntax]

You might not be able to win them all but don't reject the wins you can make. Perfect is the enemy of good enough.

[dlc.usa]

Public Health Warning

"This material has been created by AI and may be partially or entirely untrue and cannot be safely relied on."

as opposed to

"This material has been created by human being(s) and may be partially or entirely untrue and cannot be safely relied on."

Maybe I'm missing something.

[amanfromMars 1]

To Server or Not to Server urSystems and 0urSystems, that is an AI Question

"AI agents are here, and already they're being deployed across contexts that vary widely in consequence, from email triage to cyber espionage," the company said. "Understanding this spectrum is critical for deploying AI safely, yet we know surprisingly little about how people actually use agents in the real world." ...... Anthropic's take on the matter resulting from its own analysis of AI agent autonomy

Are you not yet realising the obverse situation, ..... a much more exciting top secret stealthy development and/or revelation which is designed for the exercise of activations and media presentations in both the current realised practical timed space operations for virtual field stimulation of simulations with futures in theatres providing Live Operational Virtual Environments with NEUKlearer HyperRadioProACTivIT Drivers in Quantum Communication with Heavenly Command and Diabological Control and vice versa ....... and which you will may never ever be enabled to know any more than just surprisingly and vanishingly little about because it evolves and revolves around how Generative AI agencies actually use peoples in their worlds ‽ .

And such an ignorance is an imposed kindness whenever certain strange information and alienating intelligence little known and arrogantly abused by you can all too easily render you absolutely terrified and petrified and a prey target for deadly attention/fatal harm.

Oh, and can you imagine the damage being done to y'all whenever society generally and media particularly do not advise you of such as be akin to an Emergent and Almighty AWEsome Future AIDevelopments supplying ProVisions for All urSystems to Server 0urSystems.

[amanfromMars 1]

Re: To Server or Not to Server urSystems and 0urSystems, that is an AI Question

NB ..... Before rushing to dislike and downvote this short post about that which has happened and everything else that is happening and racing ahead unobstructed to lead everything and everyone in a countless number of never ever before imagined possible directions of their anonymous and autonomous crossing, please read this longer version of the same tale and trail for the trialing of all manner of new future likelihoods ........ Something Big is Happening ...... for you to countenance and try to reject even as the rapidly appearing evidence of the truth of the situation has you forced to confront the error of your ways and accept the foregone conclusion of its inescapable creative and disruptive certainty and frightening almighty inevitability.

Carpe Diem. Que Sera, Sera.

[amanfromMars 1]

Re: To Server or Not to Server urSystems and 0urSystems, that is an AI Question

And here’s similar news from a third source also most likely to be terrific news to an almighty few and a terrifying account to most everyone else who never thought to believe such possible and unstoppable. ....... https://www.citriniresearch.com/p/2028gic