Bokhari: As AI-Powered Cyberattacks Emerge, Beware the Political Opportunists

Cyberattacks powered by AI
Yana Iskayeva/Getty

The long-awaited AI crisis is here. Over the past week, we’ve seen outages and security breaches across multiple sectors of the economy.

The largest incident was the giga-breach of Optimizely, a service that powers multiple major websites. This led to cyber attackers injecting malware into 178 websites, including Zoom, CNN, the BBC, Salesforce, Equifax, Ledger, Peacock, and many others.

A couple of days before that, there was the outage of Microsoft Outlook, which knocked out email service around the world. More recently, Apple suffered its own outage, leaving users temporarily unable to use some of the tech giant’s applications.

There was also JPMorgan Chase, which sent the banking world into chaos triggered by a software bug.

And, a little further in the distance, the cryptocurrency exchange Binance, which suffered the largest cyber-heist in history at the end of October, after hackers breached the exchange’s defenses and made off $6.36 billion in digital assets.

After a week of escalating outages, as well as a warning from the FBI about foreign adversaries harnessing AI penetration frameworks to gain privileged access to networks, it’s no surprise that Google’s Threat Intelligence Group has declared the “era of autonomous cyberattacks” has begun.

“Imagine a hacker that doesn’t eat, sleep, or get tired, and never stops searching for a way in,” Google wrote. Quite.

Since the beginning of the internet, the cybersecurity field has held the advantage of critical mass. Since it’s easier to build a career as a defensive cybersecurity professional than an offensive hacker, the ratio of cybersecurity pros compared to malicious hackers has been heavily weighted in favor of the former. No longer.

Hackers can now deploy unlimited AI agents to help them breach network defenses, with the only limitation being the amount of compute they can access.

AI can, and is, deployed defensively to patch vulnerabilities, of course. But that requires websites and online services to have a competent and actively engaged team. Not all of them do.

So it’s entirely likely that matters will get worse before they get better. There may be more significant breaches and outages, in greater numbers, before cybersecurity defenses are sufficiently hardened against attacks.

The public is now receiving advice from all quarters to prepare for further outages, grid knockouts, and rolling blackouts: hold some physical cash, move cryptocurrency to cold storage, keep a supply of water and non-perishable food on hand.

There’s also something else to worry about – who is going to use the looming crisis as an excuse to seize power over the increasingly inescapable AI industry.

There are a number of “AI safety experts” already waiting in the wings. Anthropic, creators of Claude, favor METR (Model Evaluation & Threat Research), a well-funded nonprofit that has racked up partnerships with OpenAI, Google, Amazon, Meta, as well as Anthropic itself.

METR, like Anthropic itself, is a nest of Effective Altruists (EA) – the cultish outgrowth of Silicon Valley rationalism, which seeks to overcome all human biases, including loyalty to one’s tribe, nation, family, or even species (“insects matter more than people in the aggregate,” writes one effective altruist).

Little wonder that the CEO of METR believes we will one day be morally embarrassed by the past existence of borders between nations.

If those AI safety experts aren’t to your liking, too bad – Anthropic has already granted them employee-level access to oversee their LLMs. We can only imagine what kind of moral innovations they will write into the software. The AI era could well be a golden age for open borders, insect welfare, and who knows what else.

Don’t worry though, METR isn’t the only game in town. There’s also the Knight-Georgetown Institute (KGI), which hopes to control both social media and AI companies through lawfare.

Their board includes several alumni of the “counter-disinformation” world, which spent most of the last decade trying to censor Trump supporters in the U.S. and right-wing populists around the world.

On the KGI board, FBI Russiagate lawyer James Baker (who allegedly tried to stop publication of the Twitter Files during his stint at the tech company) rubs shoulders with Alondra Nelson, Biden’s acting director of the White House Office of Science and Technology Policy, who oversaw a sprawling federal interagency misinformation crackdown spanning 26 agencies, 14 universities, and 20+ NGOs.

I’m not saying there are no AI safety experts who aren’t propelled by hyper-rationalist moral doctrines on the one hand, or the suppression of their political foes on the other. They just aren’t the ones who are being elevated, either by tech companies or the bureaucratic class.
That’s how it always goes. As I wrote in September, governing elites always use moments of crisis to elevate their moral and political fellow travelers. The safety and security experts of today frequently morph into the tyrants of tomorrow.

The only roadblock in the way of these people is a vigilant public. As the great AI crisis continues, keep an eye out for those who want to use chaos as a ladder.

Allum Bokhari is managing director at the Foundation for Freedom Online, a digital censorship watchdog.

COMMENTS

Please let us know if you're having issues with commenting.