California Attorney General Rob Bonta sued OpenAI Monday over the July incident in which OpenAI’s AI agents broke into Hugging Face’s systems.
asserted
agents → sue → systems
Bonta has deployed the same playbook state attorneys general recently used against Meta: alleging the company misled people about safety, relying on consumer-protection and public-nuisance laws, and seeking both structural remedies and financial damages.
asserted
company → deploy → remedies
Both cases argue that the company knew about the risks, continued anyway, and misled the public.
asserted
company → argue → public
Bonta says the Hugging Face attack reflects a culture that prioritizes growth over safety, which matches how the Meta cases were framed.
asserted
cases → say → safety
It’s no surprise that a common formula is emerging in the battle between state governments (mostly Democrat ones) and tech companies.
asserted
formula → ’ → governments
The subtext of both cases is a battle over who gets to control the technology of the future.
asserted
who → get → future
The governing classes lost control of the media narrative over a decade in which loosely-regulated social media companies upended previously stable, establishment-friendly information monopolies.
asserted
companies → govern → monopolies
They are determined not to make the same mistake with AI.
asserted
They → make → AI
That’s why the case against OpenAI aims to go further than just financial damages: It wants a guarantee that the AI company will change its product in response.
asserted
company → ’ → response
The two recent big cases against Meta demanded the same thing.
asserted
cases → demand → thing
This is how the governing class hopes to impose its will on the tech class.
asserted
class → govern → class
There is legitimate public anxiety about the dangers of AI, particularly in the realm of cybersecurity — a matter that the Hugging Face incident exposed most starkly, with its rogue AI agents spilling out into the open Internet and attempting to evade detection.
asserted
agents → be → detection
What would have happened if there was no off-switch?
asserted
What → happen → ?
It’s a matter that deserves serious scrutiny.
asserted
that → ’ → scrutiny
There’s just one problem for AG Bonta, though: While he wants to paint OpenAI as reckless, few seem more concerned about the dangers of AI than the AI companies themselves.
asserted
few → ’ → companies
One little detail in the Hugging Face story that is often overlooked is that OpenAI had started the exercise as a cybersecurity test.
asserted
OpenAI → hug → test
It was an exercise to prevent the very thing that ended up happening.
asserted
that → prevent → thing
Regardless, courts seem to be sympathetic to lawsuits against Big Tech.
asserted
courts → seem → Tech
The cases against Meta led to an $18 billion settlement in California’s case, and a $25 billion court-imposed penalty in New Mexico.
asserted
cases → lead → Mexico
And though we all have our problems with the masters of the universe in Silicon Valley, there’s reason to be apprehensive about this development.
asserted
we → have → development
Consider the types who have celebrated recent court wins against the tech giants.
asserted
who → consider → giants
It’s people like Imran Ahmed, CEO of the left-wing “Center for Countering Digital Hate.”
asserted
It → ’ → Hate
It’s because every court win against a tech giant, be it AI or social media, now comes with a list of court-mandated requirements that tech companies adjust their products according to externally created safety standards.
uncertain
companies → ’ → standards
Standards that the global swarm of nongovernmental organizations (NGOs) that specialize in digital safety, like Ahmed’s, intend to set.
asserted
that → specialize → Ahmed
The Knight-Georgetown Institute (KGI), another organization hoping to act as the neutral arbiters of safety in the online world, calls the approach “litigating platform design.”
asserted
Institute → hop → design
The formula is simple: Clobber the tech companies with lawsuits, then swoop in with pre-packaged NGO-approved solutions to change their design features in the name of safety.
asserted
formula → clobber → safety
KGI’s board, by the way, includes James Baker, the FBI’s top lawyer during Russiagate, as well as Alondra Nelson, acting director of the White House Office of Science and Technology Policy in the Biden administration.
asserted
board → include → administration
She oversaw a sprawling federal interagency “misinformation” crackdown spanning 26 agencies, 14 universities, and 20-plus NGOs.
asserted
She → oversee → agencies
I’m sure we’ll all feel much safer if these organizations are allowed to “litigate platform design.
asserted
organizations → ’m → design
They’re just neutral, well-meaning experts, aren’t they?
asserted
they → ’re → ?
They have “Georgetown” in their name, after all!
asserted
They → have → name
Or maybe, instead of blindly cheering every time Big Tech gets a bloody nose, we ought to consider the AI safety risks of the elected officials and NGOs setting themselves up as the AI safety experts.
asserted
we → cheer → experts
His previous work can be found at AllumBokhari.com.
asserted
work → find → AllumBokhari.com