Sure, A.I. is developing at speed, but so is the regulatory pushback

Let’s start in the U.S., where Minnesota lawmakers just passed a bill tackling the nascent problem of deepfakes. Under the bill, people could have to pay up to $10,000 in fines, or go to prison for up to five years, if they are caught non-consensually disseminating deepfake sexual images of other people, or if they spread deepfakes in the hope of influencing an election.

The bill was so popular in the Minnesota Senate that just one member voted against it, and that’s only because he wanted the fines to be higher.

Over to China, where a man has been arrested for allegedly using ChatGPT to generate a fake news article about a real, fatal train crash. The man, named Hong, generated multiple, slightly different versions of the article to help it bypass checks on a Baidu-owned blogging platform.

This is the first arrest under China’s new law covering “deep synthesis technologies.” ChatGPT—the sudden advent of which prompted the law—is supposed to be blocked in China anyway, but clearly not very effectively.

Meanwhile, two European Parliament committees have cleared their preferred version of the EU’s A.I. Act bill, which has also been heavily rejigged to take account of our new generative-A.I. reality. Digital rights organizations are delighted to see that the committees want the bill to ban facial recognition in public spaces, to forbid private “social scoring” systems, and to make sure people know when high-risk A.I. systems are making decisions about them. 

Big Tech lobbyists are grumbling that this version of the bill would prohibit too many kinds of A.I. systems—and they don’t like the fact that it would force A.I. companies to disclose which copyrighted data they used when training their models. The full parliament still needs to vote on this version of the bill next month, after which the final “trilogue” negotiations with the European Commission and member states can begin.

Zooming in on Germany and out from specifically A.I.-related matters: One of the country’s big consumer organizations just won a legal action blocking Deutsche Telekom from transmitting the personal data of its website visitors to the U.S.

Specifically, Telekom has Google Analytics embedded on its site, and the consumer advocates convinced the court that Google can’t legally transmit visitors’ IP addresses and device information back to America. Why? The same 2020 ruling from the EU’s top court that could soon see Facebook and Instagram blocked from serving European users, on the basis that U.S. surveillance practices mean Europeans’ personal data isn’t safe there. This issue is not going away anytime soon.

And one final trip, this time to Italy, where the competition regulator just opened an antitrust investigation into Apple, for allegedly abusing its dominant app-market position by putting heavier privacy restrictions on third-party app developers than on itself. 

A couple years ago, Apple put more power into the hands of users by letting them choose whether or not to be tracked by the apps on their phones. The Italian watchdog claims Apple’s consent requests are phrased to discourage tracking when it comes to third-party apps but not its own—it also accuses Apple of giving outside developers inferior data on user profiling and ad campaign effectiveness. It’s a clash between privacy and competition law that the online ad industry has been playing up ever since Apple made its pro-privacy changes. Similar investigations are also underway in Germany and Poland.

An Apple spokesperson told Fortune that its so-called App Tracking Transparency “simply gives users the choice whether or not they want to allow apps to track them or share their information with data brokers. These rules apply equally to all developers—including Apple—and we have received strong support from regulators and privacy advocates for this feature.”

More news below.

Want to send thoughts or suggestions to Data Sheet? Drop a line here.

David Meyer

Data Sheet’s daily news section was written and curated by Andrea Guzman. 

NEWSWORTHY

No raises for Microsoft's salaried staff. While certain hourly and equivalent workers at Microsoft will see salary increases this year, full-time salaried workers will not. CEO Satya Nadella informed staff of the decision in an email viewed by Fortune, citing "global macroeconomic uncertainties" as the reason. Nadella went on to say that the company will maintain the bonus and stock award budget and that the promotion budget remains steady. Also, Nadella and others on the senior leadership team will see a dip in performance-based bonuses compared to the previous year.  

More A.I. tech coming to Google Search. U.S. users can now join a waitlist for Google's Search Labs, where they will be able to test how generative A.I. will be incorporated into search results. Google says the content, which will be labeled as experimental, will be more factual than conversational. It will also include guardrails to ensure that users asking sensitive health and finance questions are directed to authoritative websites rather than simply being offered a snippet of A.I.-generated info, which is prone to errors. It’s just one of many A.I. announcements by Google yesterday as Alphabet CEO Sundar Pichai said the company is “reimagining all our products, including search.”

Encrypted DMs come to Twitter. Twitter users will have to pay to send and receive encrypted messages, as the social media platform rolls out the option for Twitter Blue subscribers, verified organizations that pay $1,000 a month for the status, or those associated with a verified organization for a price of $50 per person each month. So far, they’re only available for messages between two people and will be available for groups soon. The Verge reports that they’re not safe from all attacks like that of a malicious insider. “As Elon Musk said, when it comes to Direct Messages, the standard should be, if someone puts a gun to our heads, we still can’t access your messages,” Twitter wrote in a blog post. “We’re not quite there yet, but we’re working on it.”

ON OUR FEED

“We determined just about everything he has is fake.”

—Gamma, the owner of a Discord server that collects rare Frank Ocean recordings, in an announcement to the group about a scammer peddling tracks purportedly by the acclaimed musician. Motherboard contacted the scammer, who said they used A.I. to produce the songs, posted a sample to a leaked music forum, and sold them for about Can$13,000.

IN CASE YOU MISSED IT

This artist just made a memecoin that skyrocketed to a $77 million market cap. He mostly used GPT-4, by Marco Quiroz-Gutierrez

Airbnb CEO Brian Chesky ‘guarantees’ many bosses calling people back to the office ‘are going away to the Hamptons for the summer’, by Nicholas Gordon

Microsoft agrees to buy power from a Sam Altman-backed nuclear fusion startup that’s yet to produce any electricity, by Nicholas Gordon

Microsoft says AI can fix 3 big problems with ‘the drudgery of work,’ from ending burnout to clearing out companies’ ‘digital debt’, by Tristan Bove

IBM’s CEO says its new A.I. tools will be able to do ’30 – 50%’ of ‘repetitive’ office work after indicating his own company will pause some hiring, by Prarthana Prakash

BEFORE YOU GO

What does a peach smell like in the metaverse? Attempts to create scents in virtual reality have often been a messy undertaking involving liquids. But researchers Xinge Yu and Yuhang Li have developed a method that might be more seamless, and involves two wearable interfaces that can stick to the skin between the nose and mouth and another that attaches beneath a headset.

The interfaces use mini odor generators containing paraffin wax. When the wax is heated, it can generate multiple odors in less than two seconds. When it’s time to end the scents, a copper coil draws a magnet to the wax to cool it down. Their findings were published in Nature Communications earlier this week and included 30 scents that are easily recognizable, like rosemary, peach, and durian, MIT Technology Review reports.