Four and a half years ago, Google laid out an extraordinary plan to stop supporting third-party cookies—widely used by advertisers since the web’s early days to track what people do online—in its market-leading Chrome browser. The company said it would replace those cookies with its own “Privacy Sandbox” technology that would let advertisers target users with the most appropriate ads.
It pitched the idea as a way to “increase the privacy of web browsing,” but the plan kept hitting roadblocks—and yesterday Google all but abandoned it. What happened is pretty complicated—and it’s this complexity, reflecting the conflicting imperatives of the various players in the web ecosystem, that was Google’s undoing.
Third-party cookies—that is to say, cookies left in a browser that share or record information across multiple websites—give advertising technology companies a way to create profiles of users’ interests, which is how much of the online economy is funded. But it’s also creepy and something that many people would rather avoid.
Back in 2019, Google acknowledged the problems with third-party cookies by setting up the Privacy Sandbox project with the aim of letting users “control how their information is collected.” At the start of 2020, it said the idea was so promising that it could “sustain a healthy, ad-supported web in a way that will render third-party cookies obsolete.” Google claimed it would therefore phase out Chrome’s support for third-party cookies within two years.
As Google’s general manager of ads at the time, Jerry Dischler, explained in a Fortune commentary, the new approach would see Google itself take over the task of tracking users’ interests. Users would be put into groups based on those interests—AI would do the sorting—and advertisers would target ads at the groups, rather than individuals. This technology was called FLoC, which stood for “federated learning of cohorts.”
The problem is, Google is kind of a big deal in online advertising. So privacy advocates were skeptical that it suddenly had their interests at heart, and other ad-tech players cried foul, claiming that Google was trying to strengthen its monopoly in the space while making it harder for advertisers to track how well their ads were performing. This backlash had its most notable effect in the U.K., where the Competition and Markets Authority (CMA) opened a case, in close cooperation with the country’s privacy regulator. (The issue also plays a part in an antitrust lawsuit against Google by the State of Texas, which is due to go to trial next year, and in a series of complaints to European competition authorities.)
By mid-2021, Google was forced to announce a two-year delay to the cookie phase-out. At the start of the next year, it dropped FLoC, which digital rights campaigners had decried as something that would exacerbate online discrimination and predatory targeting. FLoC’s replacement was a new system called the Topics API, which would again see Chrome store and serve up details of users’ interests.
Meanwhile, the U.K. CMA decided that Google’s plans were actually anticompetitive, and in mid-2022 the company was forced to again delay the phase-out until the second half of 2024, to allow everyone to test its new technology and its impacts. Fast-forward to April of this year and—you guessed it—Google admitted it could only start phasing out third-party cookies in early 2025, pending CMA approval.
And now that’s not happening either. Yesterday, Google Privacy Sandbox VP Anthony Chavez published a blog post saying third-party cookies will in fact continue to be supported in Chrome, with users being given the choice to block them or not—if the CMA agrees.
Privacy campaigners at the Electronic Frontier Foundation characterized this as Google backtracking on its “privacy promise” and pointed out that rivals like Firefox and Safari block third-party cookies by default. That said, those advocates weren’t keen on Google’s proposed privacy protections anyway, saying they would still allow unacceptable tracking. The U.K. privacy regulator had also determined that the Privacy Sandbox tech had holes that violate people’s privacy rights, though it reacted to yesterday’s announcement with disappointment as “blocking third-party cookies would be a positive step for consumers.”
But the ad-tech industry is thrilled. “This is a clear admission by Google that their plan to enclose the Open Web has failed,” crowed James Rosewell from the “Movement for an Open Web”, the industry coalition that had complained to the CMA. “Their goal was to remove the interoperability that enabled businesses to work together without interference from monopolists but a combination of regulatory and industry pressure has put paid to that.”
It’s certainly a good week to be an ad-tech investor—Criteo’s share price leapt by over 12% on the news, while Pubmatic is up by over 6%. More news below.
David Meyer
Want to send thoughts or suggestions to Data Sheet? Drop a line here.
NEWSWORTHY
Meta’s illegal agreements. A National Labor Relations Board judge has decided that Meta broke the law with the separation agreements it made over 7,000 departing employees sign at the time of major layoffs in 2022. In particular, the agreements’ non-disparagement and confidentiality sections had “overly broad language.” The NLRB still has to issue a final decision in the case, but as The Register notes, upholding the judge’s decision could open the floodgates to similar cases.
Cruise setback. GM’s Cruise robotaxi unit has abandoned—or at least indefinitely postponed—plans to build its own, not-made-for-humans-to-drive vehicle, which was to be called Origin. Per TechCrunch, Cruise will just run with a next-gen Chevrolet Bolt instead. GM chief Mary Barra said this would “simplify their path to scale” while also avoiding potential regulatory pitfalls. GM will take a $583 million financial charge related to the Origin abandonment.
Intel finds chip problem. Intel says it’s identified what’s making its Core 13th and 14th Gen processors crash under some circumstances—“elevated operating voltage [that] is stemming from a microcode algorithm”—and will soon release a patch, The Verge reports. The crashes have largely been experienced by gamers and operators of game servers.
SIGNIFICANT FIGURES
$500 million
—The amount raised by Cohere in a fresh funding round that values the Canadian generative AI startup at $5.5 billion, according to Bloomberg. Cohere targets enterprises rather than consumers, building large language models that it customizes for the likes of Oracle.
IN CASE YOU MISSED IT
Wiz turns down $23 billion Google deal, says it’s aiming for IPO, by Allie Garfinkle
CrowdStrike’s CEO confronts his own crisis, by Diane Brady
The rise of the AI gadget could free us from our smartphones. We just need to find the right device, by David Meyer
Crowdstrike CEO called to testify before Congress on software update fiasco, by the Associated Press
Elon Musk is backing Trump, but Democrats won’t quit X, by the Associated Press
War and rising nationalism is expected to spur a venture capital boom in a one-time Silicon Valley ‘taboo’ sector: defense tech, by Luisa Beltran
LinkedIn rolls out puzzles in a bid to become a daily digital pit stop, by Marco Quiroz-Gutierrez
BEFORE YOU GO
Tesla delays stuff again. Elon Musk observers will be used to ever-shifting timelines, and Tesla has now delivered just that twice in two weeks. First it was the fabled robotaxi launch being delayed until October, and now it’s the Optimus humanoid robot being pushed back. As Electrek reports, Musk recently said Optimus robots would be performing tasks in Tesla plants by the end of this year, but now he’s talking about “low production for Tesla internal use next year,” followed by a “high production” rollout for other companies in 2026—“hopefully.”