POLITICO’s weekly transatlantic tech newsletter for global technology elites and political influencers.

By MARK SCOTT

Send tips here | Subscribe for free | View in your browser

WELCOME BACK TO DIGITAL BRIDGE. I’m Mark Scott, POLITICO’s chief technology correspondent, and as the deadlines (oh, so many deadlines!) creep up on me, I can’t help but think about how I never follow this advice.

OK. Let’s “complete the task” together:

— Ahead of the Munich Security Conference, the military’s use of artificial intelligence remains off limits to all but the lightest of oversight.

— Europe’s new social media rules come into full force this weekend. We are a long way from claiming, “Mission accomplished.”

— Date for the diary: The next EU-U.S. Trade and Technology Council will be held between April 4-6.

**A few months ahead of elections in Europe and the U.S., what is the bloc’s tech legacy? POLITICO’s Tech & AI Summit is a flagship convention for technology and artificial intelligence enthusiasts, bringing influencers center-stage. Interested? Join us here**

MILITARY AND AI: DO WHAT I SAY, NOT WHAT I DO

IT’S GOVERNANCE-A-PALOOZA WHEN IT COMES to artificial intelligence. The European Union is nearing completion of its new rulebook. The White House is cock-a-hoop about federal agencies implementing its executive order on AI. Everyone from the United Nations to the G7 is doubling down on safeguards for this emerging technology. But amid this global gold rush of AI standards and rulemaking, one sector is absent from the discussion: countries’ militaries and national security agencies. Almost all of these bodies are falling over themselves to harness the latest advances. But they’ve been given massive carve-outs from civilian oversight that are already having real-world consequences.

Let’s turn to two of the tragic conflicts that currently plague the world: Ukraine and Israel. In both countries, AI is entrenched in the military conflict. The likes of Palantir and ClearviewAI have offered their services to Kyiv, providing everything from data analytics analysis for drone strikes to the uncovering of potential Russian spies embedded in the local population. In Israel, Palantir, again, is working with Tel Aviv, while the country’s military is using an artificial intelligence system, dubbed Habsora, or “The Gospel,” to select targets in Gaza. The list goes on. In short, just as ChatGPT has captured the public’s imagination, militaries around the world have also become enamored with AI.

So that takes us to the inevitable question: What rules are in place to check the potentially lethal use cases of artificial intelligence? That’s particularly pressing ahead of the Munich Security Conference, an annual gathering of the West’s national security apparatus that starts Friday. Also in attendance this year: buckets of AI companies (and not just the likes of Palantir) eager to sell their wares to willing governments. Let’s leave aside the overblown chatter about deepfakes harming elections. There are enough examples of real-world AI military uses — and not just those backed by Western democracies — that figuring out what oversight should mean is now an existential question.

And on that, I have good news and I have bad news. The good news is that Western militaries acknowledge the need for safeguards. Last fall, the United States led a 52-country (voluntary) pledge on responsible military use of AI. The White House’s executive order also called on the U.S. Department of Defense to conduct studies into the risks associated with AI’s role in the armed forces. Multiple Western militaries (here and here) have backed civilian oversight, transparent data usage (under certain conditions) and a litany of other AI regulatory buzzwords — accountability, risk-based management, proportionality — aimed at giving reassurances that military AI won’t get out of hand.

“The U.S. is a global leader when it comes to putting out doctrine in terms of autonomous weaponry,”Steven Feldstein, a senior fellow at the Carnegie Endowment for International Peace and former U.S. State Department official, told me. “But where it becomes a wild west is when it comes to other countries and the capabilities they have … When you don’t have a sense of what those capabilities are, it also makes deconfliction later on, on the battlefield, a fraught endeavor.” Feldstein has a three-part plan: greater civilian oversight for U.S. military AI; international experts to monitor AI’s effect on warfare; and globally agreed limits to how militaries use this emerging technology.

And now, the bad news: All of these plans are voluntary. As much as Western governments acknowledge the need for control and accountability in how the military embraces AI, they fret that enemies — most notably Russia and China — won’t live by the same rules. When confronted with that threat (one, I should add, championed by the likes of Eric Schmidt and other China hawks), it’s not surprising current domestic and international AI governance explicitly provides national security agencies with almost complete carve-outs. Those calling for few checks on military use cases may have a point. At a recent U.N. vote outlining discussions on so-called autonomous weapons, Russia voted against the resolution (alongside India), while China (and Israel) abstained.

I can already hear the cries that, in a time of intense geopolitical rivalries, it’s naive to suggest hamstringing Western militaries when those from authoritarian countries exploit AI in ways that could eventually harm us. I get that. I have sympathy for that argument. But if the recent global AI governance discussions have shown anything, it’s that politicians — and the general public — want greater checks on the technology, and not just because of the existential threat that some believe the technology represents. If that thesis holds true, then the same wariness should apply to military applications. The hard part is threading the needle between such civilian checks while allowing national security agencies to keep us safe.

EUROPE’S NOT-SO-READY SOCIAL MEDIA PLAYBOOK

HAPPY DIGITAL SERVICES ACT DAY, EVERYONE! Well, no one is going to say that. But on Feb. 17, the European Union’s new rules for social media platforms, e-commerce players and search engines finally enter into full force. The legislation — including mandatory risk assessments to pinpoint vulnerabilities, greater accountability for how companies protect users, and regulatory fining powers of up to 6 percent of a firm’s annual turnover — already kicked in for the biggest tech giants last August. But from this weekend, many of those same provisions will apply to digital services with fewer than 45 million users across the 27-country bloc.

Before I’m accused of only focusing on the negative, let’s give Brussels its due. Unlike other parts of the world (looking at you, Washington), the bloc actually passed new rules aimed at safeguarding people from online hate, foreign interference and other nasty, illegal material. It goes further than other countries like Australia and the United Kingdom in balancing the need for transparency in how these platforms work with the ability for people to say what they want online — even if those posts, to many, are pretty distasteful. It’s also a policy wonk’s dream: The Digital Services Act is all about systemic risk assessments to find problematic content, outside audits to check for compliance, and nerdy annual reports detailing how companies are adhering to the rules.

In short, a Ministry of Truth focused on online censorship it is not. But what we’ve seen over the last six months — the time period when the rules have applied to the 22 most systemic social networks, search engines and e-commerce platforms — is a mishmash of regulatory oversight. The European Commission, which is the first among equals alongside national agencies in policing the new standards, has often run before it could walk. Limited resources and expertise have collided with real-world events, most notably the Oct. 7 Hamas attacks and its digital aftermath. Brussels has been very eager to sing the praises of its new standards without recognizing that, for now, they are more paper tiger than actual deterrence.

Let’s go through some stats. As of January, the Commission had 75 full-time officials working on social media enforcement. That equates to about three people for each of the 22 major platforms facing the toughest scrutiny. A further 50 experts are expected to be hired this year. Another 10 technicians (with 20 more soon to join) also work at the adjoined European Center for Algorithmic Transparency, a data-focused unit in charge of the heavy techie lift on upcoming investigations. It’s hard to compare one regulator’s resources to another. But to put the Commission’s personnel into context, the U.K.’s Office of Communications had about 300 people in place before London’s separate online safety safeguards came into play. Australia’s eSafety Commissioner has more almost 500 staff, although that agency’s remit is slightly different. In short, Brussels is massively understaffed for the powers it now has.

But what about national regulators, known as Digital Services Coordinators, I hear (the geekiest of) you ask? That’s true. Under the EU’s regime, these agencies also have powers to fine companies for wrongdoing. But, as of mid-February, only a third of the bloc’s 27 countries had legally appointed one of these bodies, with some (looking at you, Germany) unlikely to do so until April. These regulators meet next week via the newly-created European Board for Digital Services, a pan-European coordination group. Yet exactly how national regulators will divide their powers with the Commission is still very unclear. A potential way forward: Brussels, Dublin and Paris are already working closely, with the hope of starting joint investigations into potential wrongdoing. Let’s see where that goes.

Speaking of investigations, there’s currently a divide between the politics of Europe’s new social media rules and the reality of what effective oversight should look like. European Internal Market Commissioner Thierry Breton has gone out of his way to publicly chide companies for potential infractions. He also personally urged officials to open investigations into how X handled Oct. 7-related content, according to four people with knowledge of those discussions. FWIW, Breton got his way in December. At the top echelons of the Berlaymont building in Brussels, there’s an unrealistic expectation that the Digital Services Act can cure all social media’s ills with a flick of a regulatory switch. Such bravado, for now, does not match the reality of enforcement.

There’s a sizable number of officials — both in the Commission and within national regulators — who have a different take. These rules, they argue, will take years to apply correctly, and we should tone down the political rhetoric — or face a public backlash when things, inevitably, go wrong. These officials want to focus on procedure, due process and effective communication with the companies they now oversee. In short, what staid regulation should look like. But, annoyingly for them, that’s not how this is going to work. The political spotlight on agencies to act (and now!) is acute, especially ahead of the European Parliament election in June. That may lead to short-term wins — including the opening of investigations — before regulatory teams are in place to conduct those probes in ways that lead to good outcomes.

We definitely saw that during Europe’s last round of digital rulemaking via the bloc’s data protection overhaul in 2018: People had expected fireworks, only to become disenchanted when the legislation was more fizzle than sizzle. I hope history does not repeat itself.

BY THE NUMBERS

THE NEXT (OR LAST?) TTC

IT FEELS LIKE WE JUST DID THE USUAL SONG AND DANCE around the EU-U.S. Trade and Technology Council summit held in Washington in late January. But buckle up: The potentially ultimate version of this transatlantic gathering has been penciled in for two days between April 4-6. The date, annoyingly, has yet to be set, mostly because of the ever-changing schedule of U.S. Secretary of State Antony Blinken. But it’s either going to be April 4-5 or April 5-6. (If anyone is interested in a Digital Bridge-sponsored meet-up around the event, let me know here.)

European and American officials are eager to go out with a bang. The January version of the TTC, I’m told, was mostly to lay the groundwork to roll out completed tasks by April. That includes definitions and methodology for artificial intelligence; greater funding cooperation for telecommunications and cybersecurity projects in emerging economies; and joint research and development programs around the likes of quantum computing and sustainable trade. Will any of this come off? Officials (and you know who you are) are deep in discussions about these topics now. As always, if you want to chat, you know where to find me.

WONK OF THE WEEK

IT’S NOT OFTEN WE RETURN TO SOMEONE for “wonk of the week.” But after Elizabeth Kelly, former special assistant for economic policy at the White House’s National Economic Council, was appointed director of the newly created U.S. AI Safety Institute, the timing felt right.

The Yale and Oxford graduate is in charge of the body, nominally housed inside the National Institute of Standards and Technology, working with academics, government officials, companies and civil society groups to build so-called trustworthy AI.

“Elizabeth has been a driving force behind President Biden’s AI executive order to advance safe, secure, and trustworthy AI,” said Bruce Reed, the president’s deputy chief of staff, who oversaw last year’s federal decree.

THEY SAID WHAT, NOW?

“Finalizing the remaining work needed to implement the two-pillar agreement on the revision of the international tax rules is a top priority,” according to an internal EU document discussing the proposed global tax overhaul, including Pillar 1 (a global levy on the world’s 100 largest companies) and Pillar 2 (a minimum worldwide corporate tax rate of 15 percent). “The G20 should urge … that the Multilateral Convention (MLC) for Pillar 1 is ready for signature by the end of June 2024 at the latest and call for its swift ratification.”

WHAT I’M READING

— A team from the St. Gallen Endowment for Prosperity and Trade created a chatbot that allows you to ask questions solely about the EU’s Artificial Intelligence Act. Versions for Chinese and U.S. artificial intelligence rulemaking will be coming soon.

— The Integrity Institute’s Jeff Allen explains why Instagram’s search function remains so toxic, and offers an alternative to make it better. Take a look here.

— The effects of generative AI on this year’s global election cycle are still unknown and unquantifiable. But the technology may still be felt worldwide, argues Valerie Wirtschafter for the Brookings Institution.

— The U.S. State Department provides an update on how Russian-based disinformation is spreading across Africa. It’s biased, but worth a read.

— Sitra, the Finnish Innovation Fund, outlines its vision for how Europe can compete with others globally in the so-called data economy over the next five years. More here.

— Germany’s federal government published an international strategy for digital policy, including the promotion of rights online and free-flow of data. Read it here.