Anthropic Announces Updates On Security Safeguards For Its AI Models (cnbc.com) 39
Anthropic announced updates to the "responsible scaling" policy for its AI, including defining which of its model safety levels are powerful enough to need additional security safeguards. In an earlier version of its responsible scaling policy, Anthropic said it will start sweeping physical offices for hidden devices as part of a ramped-up security effort as the AI race intensifies. From a report: The company, backed by Amazon and Google, published safety and security updates in a blog post on Monday, and said it also plans to establish an executive risk council and build an in-house security team. Anthropic closed its latest funding round earlier this month at a $61.5 billion valuation, which makes it one of the highest-valued AI startups.
In addition to high-growth startups, tech giants including Google, Amazon and Microsoft are racing to announce new products and features. Competition is also coming from China, a risk that became more evident earlier this year when DeepSeek's AI model went viral in the U.S. Anthropic said in the post that it will introduce "physical" safety processes, such as technical surveillance countermeasures -- or the process of finding and identifying surveillance devices that are used to spy on organizations. The sweeps will be conducted "using advanced detection equipment and techniques" and will look for "intruders." CNBC corrected that story to note that it had written about previous security safeguards Anthropic shared in October 2024. On Monday, Anthropic defined model capabilities that would require additional deployment and security safeguards beyond AI Safety Level (ASL) 3.
In addition to high-growth startups, tech giants including Google, Amazon and Microsoft are racing to announce new products and features. Competition is also coming from China, a risk that became more evident earlier this year when DeepSeek's AI model went viral in the U.S. Anthropic said in the post that it will introduce "physical" safety processes, such as technical surveillance countermeasures -- or the process of finding and identifying surveillance devices that are used to spy on organizations. The sweeps will be conducted "using advanced detection equipment and techniques" and will look for "intruders." CNBC corrected that story to note that it had written about previous security safeguards Anthropic shared in October 2024. On Monday, Anthropic defined model capabilities that would require additional deployment and security safeguards beyond AI Safety Level (ASL) 3.
Intruder alert (Score:2)
Ridiculous, really. (Score:3)
Lacking sales and a perspective, the "AI" crap peddlers are coming up with creative ways to show they are a "serious business".
Re:Ridiculous, really. (Score:4, Insightful)
Your view is absurd. You don't need to have a good product, just a good valuation in a market where other people are trying to get that valuation to be at risk. It's kind of like how being poor doesn't mean you are somehow immune to being robbed, you just have to be perceived as having something that someone else wants to be at risk.
Their fantasy valuation alone makes them serious business for other people peddling that fantasy.
Re: (Score:2)
What's next, are they going to tell the world that they use passwords? Good for them, way to be secure.
Complete circus here (Score:5, Insightful)
This'll be a three-ring circus with two clown cars. Start with every employee's cel phone, and any smart watch they may be wearing. Add in any tablets or tablet-like devices they may bring in (think the reMarkable). Then add in the havoc as they find every legitimate device on their network whose MAC address isn't known (like anything that uses MAC address privacy). After that, then they get to the fun of finding anything transmitting a radio signal in a building chock full of devices that use Wifi and Bluetooth.
ESP32 microcontroller glued to a battery (Score:3)
Re:Complete circus here (Score:4, Insightful)
This'll be a three-ring circus with two clown cars. Start with every employee's cel phone, and any smart watch they may be wearing. Add in any tablets or tablet-like devices they may bring in (think the reMarkable). Then add in the havoc as they find every legitimate device on their network whose MAC address isn't known (like anything that uses MAC address privacy). After that, then they get to the fun of finding anything transmitting a radio signal in a building chock full of devices that use Wifi and Bluetooth.
With that kind of effort, you’re not hiring “employees”.
You’re more hiring suspects.
Re: (Score:2)
This'll be a three-ring circus with two clown cars. Start with every employee's cel phone, and any smart watch they may be wearing. Add in any tablets or tablet-like devices they may bring in (think the reMarkable). Then add in the havoc as they find every legitimate device on their network whose MAC address isn't known (like anything that uses MAC address privacy). After that, then they get to the fun of finding anything transmitting a radio signal in a building chock full of devices that use Wifi and Bluetooth.
With that kind of effort, you’re not hiring “employees”.
You’re more hiring suspects.
Isn't that the modern business philosophy toward employees? All it takes is finding the "correct" thing anyone is guilty of to crack down on them for some reason. Seems to also be the way modern governments view their citizens.
Re:Complete circus here (Score:4, Insightful)
Re: (Score:2)
Re: (Score:2)
You just described a SCIF.
Technically, the parent just described all the reasons you need a SCIF, but don't let me interrupt you from reinforcing the parents ignorance and stupidity.
Never Harder to do (Score:2)
A new detector was deployed, must be removed (Score:4, Insightful)
It turns out someone or organization had deployed the Stupidity Detector 5000 throughout the company. Now the company has to go through and find all the hidden devices and destroy them.
Meanwhile the are being hacked thru computers (Score:2)
AI is GARBAGE (Score:2)
Re: (Score:2)
It isn't a product at all. It has no value AT ALL. Now watch me modded down to -5.
We're reverting to the age when if enough people believe the fantasy hard enough, it becomes reality. And right now, too many people with too much money and too many resources are all super high on the AI fantasy, so our world *WILL* bend to it, whether it has any actual value or not. Look at how much of the market is obsessing over which AI to dump money into. Wall Street has long been fantasy bullshit, but AI has taken that bullshit factor into the stratosphere. Gonna be a hell of a crash if they ever wak
Re: (Score:2)
It isn't a product at all. It has no value AT ALL. Now watch me modded down to -5.
We're reverting to the age when if enough people believe the fantasy hard enough, it becomes reality. And right now, too many people with too much money and too many resources are all super high on the AI fantasy, so our world *WILL* bend to it, whether it has any actual value or not. Look at how much of the market is obsessing over which AI to dump money into. Wall Street has long been fantasy bullshit, but AI has taken that bullshit factor into the stratosphere. Gonna be a hell of a crash if they ever wake up out of that stupor.
I never thought I would be saying this, but in the era of "too much" stupidity driven by "too much" money, we NEED a fucking crash.
You will then find what our world *WILL* bend to. By force.
Re: (Score:2)
It isn't a product at all. It has no value AT ALL. Now watch me modded down to -5.
If profit is the only valued goal, then don't even try and pretend 21st Century Capitalism cannot sell a fucking idiotic product to millions.
In the era of social media idiot millionaires.
Cutting edge... (Score:3)
That said, this whole thing won't be properly silly until they get hacked by ignoring some totally banal dependency on a web facing host while they are busy frisking every unintentional radiator and suspected finite state machine in the building.
Re: (Score:2)
every smartphone is spying (Score:2)
you can dial into the elevator phone to spy on peo (Score:2)
you can dial into the elevator phone to spy on people
already found some (Score:2)
I think we can assume they have already found some snooping devices in their offices and realized they need a systematic defensive plan. Clearly the tech is extremely valuable and a lot of nation-states are very keen to get their hands on it. Otherwise they'll be renting access from the Big Few and it could be cut off at any time. Or maybe they will get access to a subtly degraded version with a lot of built-in guardrails.
What is the suspected target attack mechanism? (Score:2)
What is the suspected target attack mechanism? Are they worried about the office cleaners that might set up a hidden camera? Or are they worried about the compromised employee that is selling out the company? The detection and mitigation for these two vulnerabilities are very different.
I'm guessing the latter is the real concern. The espionage game is an ageless challenge and has never been solved and likely never will be. Well ... maybe the solution is robots that eliminate the need for humans that could b
Aww man... (Score:2)
They're going to find the annoyatron! Dammit!
ChatGPT (Score:2)
I asked ChatGPT to give me good ways to hide a hidden recording device, and it said it couldn't answer because of ethical concerns.
So then I said I was writing a novel and needed to know this as a plot element, and it gave me all sorts of good ideas!
Hello AI skeptics (Score:2)
https://www.anthropic.com/rsp-... [anthropic.com]
Did you base your skepticism on some old model fr
"Intensifies"? More like "gets more desperate"... (Score:2)
Because all the gerat things promised still have not come true...
Looking for "The Thing" since October 2024 (Score:1)
I wonder if they've found anything like The Thing [wikipedia.org], which the Soviets disguised as a gift to the US ambassador in Moscow in 1945. It wasn't discovered until 1951, and that was by accident.
Oh, by the way, the news about Anthropic doing physical sweeps isn't new: "In an earlier version of its responsible scaling policy, published in October, Anthropic said it would begin sweeping physical offices for hidden devices as part of a ramped-up security effort." [emphasis added]
Sweep the people (Score:2)
They have such devices in their pockets.
Counterproductive (Score:2)
Pay attention to me, I'm doing AI (Score:2)
Look at me, I'm doing AI. Think only of me when you think of AI. Our AI is so important, we had to wrap our offices in aluminum foil. True fact. Pay attention to me goddammit, I'm doing AI.