Google has monopolized the technology used to buy and sell online display ads by restricting or eliminating the choices of its customers—both website publishers and advertisers—the U.S. Justice Department alleged at a federal antitrust trial.

Antitrust enforcers have sought to illustrate how the Alphabet Inc. unit’s complex ad ecosystem works and the ways in which the company allegedly manipulated the features of its products and the rules of its auctions to its own benefit. Over the last two weeks at the Virginia trial, they accused Google of abusing its market power in three areas: sell-side tools used by websites, called ad servers; advertising exchanges; and buy-side tools used by advertisers known as ad networks.

Website publishers use an ad server to manage space available for sale. The ad server acts as the brain for the website, keeping track of the minimum bids a publisher is willing to accept, what has been sold and for how much. The Justice Department estimates that Google’s ad server controls 87% of the U.S. market and 91% of the market globally.

Ad exchanges control the auctions that match website publishers with advertisers. Google operates the largest exchange, known as AdX, later rebranded as Google Ad Manager. The Justice Department estimates that Google’s ad exchange controls 47% of the U.S. market and 56% globally. Other popular ad exchanges include Pubmatic Inc., Index Exchange, and Magnite Inc.

Sophisticated advertisers use software known as a demand-side platform to manage their ads and help determine which ad exchanges to bid on and for how much. Google operates a demand-side platform that can bid on an ad exchange.

Advertisers also use ad networks, which take over most of the decision-making process like where to place ads and what to bid, and are most often used by smaller companies. The Justice Department alleges that Google’s network, Google Ads, controls 88% of the U.S. market and 87% globally.

Antitrust enforcers alleged that Google gave special access and privileges to its own ad products to encourage both advertisers and websites to spend only through its services.

Google has argued that the Justice Department’s case misunderstands the dynamics, pace of innovation and competitive landscape within the online advertising market. Advertisers have multiple choices for where to buy ads, the company said, including Amazon.com Inc., Meta Platforms Inc.’s Facebook and Instagram, as well as Amazon.com Inc. and ByteDance Ltd.’s TikTok.

The company also said that many of the Justice Department’s contentions mischaracterize how the technology operates. Changes to the ad tech platform were intended to improve the product, Google said. The limitations on rivals’ access were in place to reduce spam and ad fraud or help advertisers have better control over where ads appeared, the company said.

Tying products

As one advertising executive wrote in a 2017 email to her boss: “The value in Google’s ad tech stack is less in each individual product but in the connections across all of them.”

Advertisers using the Google Ads network are only permitted to place bids through Google’s own exchange, AdX, with a few limited exceptions. That gives AdX a significant volume of ads. In 2020, for example, Google Ads sent in bids for 18 million ads sold through AdX, but only about 3 million to 4 million to third-party exchanges. The tight connection between the products requires any websites that want Google Ads advertising to use AdX.

Similarly, certain functions of Google’s ad exchange, such as real-time bidding, are only available to publishers that use its ad server, DFP, according to the Justice Department.

“Customers are essentially forced to use DFP to get access to AdX,” said Rosa Abrantes-Metz, the Justice Department’s economic expert. Likewise, “AdX is the only channel to reach Google Ads in full.”

Popular websites testified that they felt compelled to use Google’s ad server product because of its exclusive access to Google Ads.

“I felt like they were holding us hostage,” said Stephanie Layser, formerly a top executive at News Corp. in charge of the media company’s use of advertising technology. Google’s ad server technology, developed decades ago, was “slow and clunky,” she said, but News Corp. estimated it might lose as much as $9 million a year if it moved to a different server because it would lose access to Google Ads.

Companies that work with advertisers said they had to use Google’s ad exchange, even though it charges higher fees than others, to ensure they had enough access to website inventory.

Google has argued that it has no obligation under the law to make its products work with those offered by rivals. The requirement that Google Ads bid almost exclusively through AdX helps the company better manage spam and ad fraud, company employees testified.

In 2015, Google launched a program called AdWords Bidding that allowed more exchanges to bid on Google’s inventory. But it was limited to ad retargeting campaigns, by which marketers reach out to users who have visited their website in the past.

Opening up the platform to allow more exchanges to bid on Google’s inventory required a huge lift from the company’s internal engineering teams to work through others’ problems with quality, testified Nirmal Jayaram, a Google engineer who worked on the company’s advertising tools for marketers.

First look

Before 2015, online display ads were sold using a “waterfall” method. A website’s ad server would ask each advertising exchange for bids sequentially. If the first exchange had an advertiser willing to pay a website’s minimum bid, the process ended. Subsequent exchanges only got to bid if the first one didn’t want a particular ad impression, even if they would have been willing to pay more.

The majority of websites used Google’s ad server, which would automatically call on Google’s ad exchange to bid first, according to the Justice Department. That advantage allowed Google to win most often. Even websites that used a different ad server tended to seek out Google’s exchange first because of its exclusive access to advertisers using Google Ads.

Facebook’s Brian Boland, who helmed the social network’s ad tech tools for a decade, likened this to Google getting to pick the best apples out of a crate before any other exchange.

Google was aware of its advantage. “It’s strategic for us to have the AdServer being the decision maker to ensure” Google’s advertiser product “has first look access,” Eisar Lipkovitz, who headed the company’s display ad business from 2014 to 2019, wrote in an email exchange with other Google colleagues.

Gabriel Weintraub, an expert who testified for the Justice Department, estimated that Google’s first look decreased the share of bids that rival exchanges saw by about 25%.

Google said the DOJ’s descriptions of how the ad server worked mischaracterize the technology and that it was technically possible for websites to move another exchange in front of Google’s AdX in the waterfall. Few publishers used this technique, the company said, because they wanted to seek bids from the advertisers who had already chosen to work with Google.

AdMeld acquisition

In 2010, Google became worried about competitors known as yield management systems, which helped websites analyze historical ad performance and determine what order to seek bids from the different ad exchanges. In an internal presentation, Google employees said that the rival tools prevented Google’s advertising exchange from seeing all of the available publisher inventory.

At the time, some publishers were more comfortable with these tools because they still hadn’t wrapped their heads around some of the new features Google was offering, including real-time bidding, according to Neal Mohan, a leader in Google’s display advertising business who later became chief executive officer at YouTube. Mohan likened yield managers to DVDs in a world with modern-day video streaming, framing the tools as outdated tech.

But because publishers still used them, Google was interested in acquiring a company that made them. According to documents shown in court, in 2008, Mohan suggested “picking up the one with the most traction and parking it somewhere.”

Google eventually decided to buy a leading yield manager, AdMeld, for more than $400 million in 2011. The search giant incorporated some of AdMeld’s technology into its own ad exchange, and then shut down the service in 2013.

Last look

DOJ expert Abrantes-Metz said Google’s purchase of AdMeld was a classic “killer acquisition”—buying a potential rival and deprecating the features that threatened its own product.

By 2015, websites had become frustrated with Google and started using an alternate technology, known as header bidding, which moved the bidding process outside of the ad server. With header bidding, a publisher would add code to the website so that an auction would take place within the browser as a page was loading. This allowed websites to simultaneously seek bids from all exchanges and pick the one that would pay the most.

After the auction had ended, the website would send the information to the ad server. Because Google’s ad server is tied into its exchange, though, the company had an opportunity to decide whether it wanted to preempt the winner and take the ad for itself. Lawyers for the Justice Department argued that this allowed Google to win ads without having to directly compete against other exchanges.

The informational advantage also allowed Google to adjust its own bids downward since it had more data on the highest bids coming from other exchanges, the DOJ said.

Abrantes-Metz, the Justice Department expert, likened it to having other exchanges take part in a sealed bid auction, but letting Google open the envelope at the end.

Google acknowledged this advantage in its internal documents. Google’s tools “have visibility in to remnant price before they submit bids (commonly referred to in the market as ‘last look’),” one presentation said. But the company’s lawyers also argued that rival ad exchanges could have built integrations with its ad server that they chose not to do because it was costly.

“Both publishers and exchanges have very strongly complained about the fairness of” last look, another Google employee wrote in a 2016 email.

Weintraub estimated that Google’s last look advantage increased the amount of ad spend on its platform by $473 million a year.

Google argues that websites could avoid giving its exchange a last look, but that most publishers opted to allow it to bid because of the greater demand in Google Ads. The company also said that when Google’s exchange gained a “last look,” that led to higher revenue for publishers.

Revenue share

Google’s advertising exchange, AdX, charges a 20% fee on winning bids. But in 2014, Google introduced a mechanism, called Dynamic Revenue Share, that would allow AdX to vary the fee if that would help an advertiser to win. The system would keep track of when it had given an advertiser a discounted fee on a particular bid and make up the difference in a later auction.

The search giant was only able to implement DRS because of its last look advantage, Weintraub said, since Google knew what price it needed to beat in order to win. He estimated that it increased ad spending on Google’s exchange by about $162 million a year.

“Because they could see all the bids, they could adjust the rev share at the end,” Brian O’Kelley, the former CEO of AppNexus, a rival ad exchange that was later purchased by AT&T and then Microsoft Corp. “Because they were at the end of the process, because they owned the ad server, they could win that impression.”

Google argues that the mechanism helped publishers because they made more money from their ad inventory. In one 2019 experiment the company conducted internally, it found that the DRS led to an increase of as much as 4% in publisher revenue.

Pricing rules

After the introduction of header bidding, Google discovered that websites would often give AdX a higher minimum price to compensate for its last look advantage.

In 2019, Google introduced Unified Pricing Rules, which eliminated the ability of websites to give Google’s exchange a higher minimum price than other exchanges.

Website publishers were very unhappy with the change. At an April 2019 meeting with top publishers where Google announced the new rules, several websites expressed concern that Google was eliminating a key tool they used to control how exchanges bid on ads.

The new rules “stopped our ability to set floors as we would like,” Michael Wheatland, an executive with the UK’s Daily Mail newspaper, testified. After the rules were adopted, the newspaper saw three times as much of its ad inventory sold through AdX, he said.

The new rules hurt publishers by limiting their ability to distinguish between different exchanges, DOJ expert Weintraub said. He estimated the rules pushed as much as $221 million in advertising spend each year to AdX.

Google disputes Weintraub’s assessment of the impact of the rules, attributing them to a separate change in bidding order.

2024 Bloomberg L.P. Distributed by Tribune Content Agency, LLC.

Alaska Airlines said it grounded its flights in Seattle briefly on Sunday night due to “significant disruptions” from an unspecified technology problem that was resolved by about 10 p.m. local time.

In comments from its account on X to customers complaining of delays and problems with the airline’s app and website, the carrier apologized for the delays. It later reported that the problem had been resolved. The exact reason for the disruptions was unclear.

“If you are traveling today, please check your flight status before leaving for the airport. If your schedule allows, please change or cancel your flight,” the airline said in a statement on its home page. “We apologize for the inconvenience and are working quickly to resolve the issue.”

It has been a rough few weeks for people traveling through Seattle, a busy hub for Alaska Airlines and other major carriers.

Last week, the operator of Seattle-Tacoma International Airport, the main hub for Alaska Airlines, said hackers were demanding $6 million in bitcoin for documents they stole during a cyberattack in August and then posted on the dark web. The Port of Seattle, which owns and runs the airport, said it had decided not to pay.

The airport has been recovering from the attack, which began Aug. 24, a busy time days before the Labor Day holiday weekend.

© 2024 The Associated Press. All rights reserved. This material may not be published, broadcast, rewritten or redistributed without permission.

Last month, OpenAI came out against a yet-to-be enacted Californian law that aims to set basic safety standards for developers of large artificial intelligence (AI) models. This was a change of posture for the company, whose chief executive Sam Altman has previously spoken in support of AI regulation.

The former nonprofit organization, which shot to prominence in 2022 with the release of ChatGPT, is now valued at up to US$150 billion. It remains at the forefront of AI development, with the release last week of a new “reasoning” model designed to tackle more complex tasks.

The company has made several moves in recent months suggesting a growing appetite for data acquisition. This isn’t just the text or images used for training current generative AI tools, but may also include intimate data related to online behavior, personal interactions and health.

There is no evidence OpenAI plans to bring these different streams of data together, but doing so would offer strong commercial benefits. Even the possibility of access to such wide-ranging information raises significant questions about privacy and the ethical implications of centralized data control.

Media deals

This year, OpenAI has signed multiple partnerships with media companies including Time magazine, the Financial Times, Axel Springer, Le Monde, Prisa Media, and most recently Condé Nast, owner of the likes of Vogue, The New Yorker, Vanity Fair and Wired.

The partnerships grant OpenAI access to large amounts of content. OpenAI’s products may also be used to analyze user behavior and interaction metrics such as reading habits, preferences, and engagement patterns across platforms.

If OpenAI gained access to this data, the company could gain a comprehensive understanding of how users engage with various types of content, which could be used for in-depth user profiling and tracking.

Video, biometrics and health

OpenAI has also invested in a webcam startup called Opal. The aim is to enhance the cameras with advanced AI capabilities.

Video footage collected by AI-powered webcams could translate to more sensitive biometric data, such as facial expressions and inferred psychological states.

In July, OpenAI and Thrive Global launched Thrive AI Health. The company says it will use AI to “hyper-personalize and scale behavior change” in health.

While Thrive AI Health says it will have “robust privacy and security guardrails,” it is unclear what these will look like.

Previous AI health projects have involved extensive sharing of personal data, such as a partnership between Microsoft and Providence Health in the United States and another between Google DeepMind and the Royal Free London NHS Foundation Trust in the United Kingdom. In the latter case, DeepMind faced legal action for its use of private health data.

Sam Altman’s eyeball-scanning side project

Altman also has investments in other data-hungry ventures, most notably a controversial cryptocurrency project called WorldCoin (which he cofounded). WorldCoin aims to create a global financial network and identification system using biometric identification, specifically iris scans.

The company claims it has already scanned the eyeballs of more than 6.5 million people across almost 40 countries. Meanwhile, more than a dozen jurisdictions have either suspended its operations or scrutinized its data processing.

Bavarian authorities are currently deliberating on whether Worldcoin complies with European data privacy regulations. A negative ruling could see the company barred from operating in Europe.

The main concerns being investigated include the collection and storage of sensitive biometric data.

Why does this matter?

Existing AI models such as OpenAI’s flagship GPT-4o have largely been trained on publicly available data from the internet. However, future models will need more data—and it’s getting harder to come by.

Last year, the company said it wanted AI models “to deeply understand all subject matters, industries, cultures, and languages,” which would require “as broad a training dataset as possible.”

In this context, OpenAI’s pursuit of media partnerships, investments in biometric and health data collection technologies, and the CEO’s links to controversial projects such as Worldcoin, begin to paint a concerning picture.

By gaining access to vast amounts of user data, OpenAI is positioning itself to build the next wave of AI models—but privacy may be a casualty.

The risks are multifaceted. Large collections of personal data are vulnerable to breaches and misuse, such as the Medisecure data breach in which almost half of Australians had their personal and medical data stolen.

The potential for large-scale data consolidation also raises concerns about profiling and surveillance. Again, there is no evidence that OpenAI currently plans to engage in such practices.

However, OpenAI’s privacy policies have been less than perfect in the past. Tech companies more broadly also have a long history of questionable data practices.

It is not difficult to imagine a scenario in which centralized control over many kinds of data would let OpenAI exert significant influence over people, in both personal and public domains.

Will safety take a back seat?

OpenAI’s recent history does little to assuage safety and privacy concerns. In November 2023, Altman was temporarily ousted as chief executive, reportedly due to internal conflicts over the company’s strategic direction.

Altman has been a strong advocate for the rapid commercialization and deployment of AI technologies. He has reportedly often prioritized growth and market penetration over safety measures.

Altman’s removal from the role was brief, followed by a swift reinstatement and a significant shakeup of OpenAI’s board. This suggests the company’s leadership now endorses his aggressive approach to AI deployment, despite potential risks.

Against this backdrop, the implications of OpenAI’s recent opposition to the California bill extend beyond a single policy disagreement. The anti-regulation stance suggests a troubling trend.

OpenAI did not respond to The Conversation’s request for comment before deadline.

This article is republished from The Conversation under a Creative Commons license. Read the original article.

When tickets for Green Day’s 2025 Australian tour went on sale, fans joined a queue—a ritual that has been practiced for decades on footpaths, on phones, and now online.

But as Green Day fans reached the purchase point, the price varied. For some, a seated ticket rose as high as A$500.

A week earlier, tickets to the Oasis reunion tour in the United Kingdom—arguably the hottest ticket in the world—rose by hundreds of pounds while on sale.

Ticketmaster calls this “In Demand” pricing. It’s an instance of what is more broadly known as dynamic pricing.

What is dynamic pricing?

Dynamic pricing is well established in tourism and air travel. In these markets, supply is fixed—the number of hotel rooms and plane seats—but demand has peaks and troughs. Prices are adjusted to maximize profit and to shape consumer behavior.

However, there are significant differences when it comes to music concert tickets.

Consumers see accommodation and transport prices up front, before committing to a “queue.” When it comes to the current practice for live music dynamic pricing, costs aren’t seen until they reach the front of the queue. There, consumers are presented with two numbers: a price and a timer counting down.

And unlike accommodation and transport services, each concert by a major touring artist (let alone Oasis) is a much more limited commodity.

How much is a ticket?

In 1964, Australians paid up to $3.70 ($63 in today’s terms) to see the world’s hottest act, The Beatles.

The relative price of concert tickets has changed with the economic and cultural role of live music.

For decades, concerts were primarily a way to promote record sales, with physical albums being the main source of revenue. Since the digital de-valuing of recorded music, live performance has become increasingly essential for the industry.

So ticket prices have risen, and with them the drive for everyone involved to get their cut.

Even better than a cut is control.

The live music market

Will dynamic pricing be accepted or rejected by the market? This question assumes a competitive marketplace. In today’s live music sector, competition is on the wane.

Australia’s live music market is dominated by three key players, two of which are owned by foreign multinationals, with Live Nation Entertainment (which owns Ticketmaster, Moshtix, and majority-shares in many Australian festivals and venues) dominating much of the market for international touring acts.

Despite the crisis in local and grassroots live music, Live Nation Entertainment has reported record revenues for the last two years.

Live Nation’s subsidiary Ticketmaster says dynamic pricing enables “artists and other people involved in staging live events to price tickets closer to their true market value.”

The promoter of Green Day’s tour is Live Nation, which manages artists, owns Ticketmaster and also controls some of the venues, and has similar interests in the Oasis tour—extending even to security. In the case of Green Day, Live Nation controls most of the supply chain.

In such a concentrated market, the test for a new pricing model is less about what consumers will choose, than what they can tolerate. Is this “true market value,” or an abuse of market power?

Could the government intervene?

Ticketmaster’s dynamic pricing is under investigation by fair trading authorities in Europe and the UK. The United States government is suing Live Nation-Ticketmaster for misuse of monopoly power.

Australian authorities are yet to announce any equivalent actions.

Justifications for market intervention are to sustain an industry—because of its economic or social value—and to extend equitable access to important goods and services. Utilities, insurance, health, and education fall into this category.

But what about culture?

Shared group experiences are becoming increasingly rare in our atomized, algorithm-driven world. Concerts contribute to social cohesion and build communities that otherwise would have fewer opportunities to meet.

If such experiences become prohibitively expensive, many will be excluded from one of the sweetest fruits of the social contract. Rather than “market value,” this would essentially amount to a failure of policy.

Stadium concerts have been on the rise in Australia for years even as other events have struggled, with the live music sector becoming more top-heavy as it consolidates around major events; a relatively recent phenomenon.

The use of dynamic pricing for Green Day tickets imports a new twist into the Australian music market, and fans are understandably upset.

The decision reflects a band that has moved a long way from their DIY origins playing all-ages communes and squats in the East Bay area, and a music industry that looks nothing like the early 1990s, when a band like Green Day could work their way up the “toilet circuit” to become a major label smash hit.

Will dynamic pricing stick—or will it force attention to the structural issues in our live music sector? With today’s news that Dua Lipa tickets will also be subject to the practice, it seems that we may be stuck with it. For now.

This article is republished from The Conversation under a Creative Commons license. Read the original article.