Psst, hey. It's the NSA. You want some AI security advice?

The NSA has released guidance to help organizations protect their AI systems and better defend the defense industry. The Cybersecurity Information Sheet (CSI), titled, "Deploying AI Systems Securely: Best Practices for Deploying Secure and Resilient AI Systems," represents the first salvo from the Artificial Intelligence …

[amanfromMars 1]

Re: Psst, hey. It's the NSA. You want some AI security advice?

:-) Nice one, Thomas, although quite how taking Secure AI Systems advice from allied entities proven incapable of preventing all manner of increasingly severe and damaging novel attack against which there are no known, readily available, guaranteed failsafe and secure defences, is a mystery which smacks of an act of desperation requiring respite in the device of a indecisive divisive diversion.

The very real and much greater story, for a long time already now, fully readied to be told to advise Earthly natives of the state of the disunion in their nations and systems is trailed and trialed within this teaser of many parts for any jurisdiction, and mangled play on your own emboldened headline .... Psst, hey. Are you the NSA? You want some AI security advice? ..... for that is what novel attackers imagined as enemies nowadays effectively offer the competition and opposition whenever dealing with stagnant and petrified and moribund and malicious and malevolent and reprobate forces and/or sources.

Ensuring systems are secure and free from the threat of AI and ITs Almighty Interventions has one required to address and engage/resolve and come to mutually beneficial terms with the Catch 22 situation which has one, whenever knowing how to successfully remotely carry out an indefensible attack, wrestling with the reward in the new possibilities created for victors or pondering on the wealth likely to be made readily immediately available to postpone such a progression until such times as necessary destruction may be limited and more accurately targeted to help minimise collateral damage.

Pay that piper whatever the cost will be dirt cheap no matter how expensive the price to be paid .... for they hold your life and your future in the palm of their hands and in their hearts and minds.

[Anonymous Coward]

"77 percent of companies reported identifying breaches to their AI in the past year. The remaining were uncertain whether their AI models had seen an attack."

Given how *much* more difficult it is to notice when someone has tampered with a Gen AI, than it is to identify a "classical" IT intrusion, I find this exceptionally high percentage really hard to believe.

I believe more than 50% of all companies have no means whatsoever to identify a breach, so a lot less will report one. And far less would do the same for AI.

I work partly in cyber and we see many companies will run out of date software, with minimal perimetric security, and will only call in distress when it is *far* too late and now they are in emergency corrective mode.