Coverage of the widespread adoption of AI software has made it seem as though it represents the latest Wild West frontier of tech innovation. That impression was reinforced in the recent corporate melodrama surrounding the abrupt dismissal, and later rehiring, of OpenAI CEO Sam Altman; surely this must be another example of nimble startups and brash entrepreneurs moving fast and breaking things.

But the actual moral of the Altman saga was the opposite: When the dust settled on all the boardroom intrigue, major market players—notably, the software monopoly platform Microsoft—were better positioned than ever to be calling the shots. Microsoft, under the leadership of CEO Satya Nadella, leveraged its exclusive contracts and control over essential cloud-computing infrastructure into a stronger hold over the company’s long-term strategic direction.

On paper, Microsoft had no board seats or formal power over the company’s decisions. Under OpenAI’s structure, a nonprofit charity called OpenAI, Inc. owns a holding company that in turn holds a majority stake in the for-profit OpenAI, LLC, with Microsoft joining the holding company as a 49 percent investor in the for-profit arm. Altman ostensibly answered only to a governing board pursuing broader dictates. That governing board had pledged to ensure the adoption of AI capabilities to improve general welfare, rather than to maximize shareholder returns. Indeed, he original call to let Altman go was rooted in disagreements over Altman’s determination to commercialize AI applications with all deliberate profit-making speed.

But in the endgame of the Altman blowup, it was the mission of the nonprofit board, not the accelerated adoption strategy, that proved the central casualty. Nadella quickly, and deftly, used Microsoft’s critical leverage over essential computing resources to lure the just-fired Altman into an inviting executive perch at Bill Gates’s software empire. To sweeten the deal, Nadella also pledged to Altman that most of his OpenAI staff could come along with him. As OpenAI’s board members pondered a mass exodus of its engineering staff, they reversed their decision by the time the weekend was over. Altman got his job at OpenAI back, along with a restructured, more Microsoft-friendly board of directors.

But Microsoft won far more than a boardroom battle here. Its enhanced role in the governance of OpenAI means that it will continue to enjoy exclusive access to the company’s technology, while playing a greater role in shaping the development and future profitability of generative AI. Even though OpenAI remains an independent entity on paper, all the relevant players understand that the software giant is really calling the shots.

This is precisely why Nadella’s power play set off instant alarms in antitrust circles. His proposed move to hire the entire OpenAI staff appeared to represent a de facto acquisition of OpenAI. Would antitrust regulators challenge a merger that involved not the acquisition of a company, nor its assets, but rather its staff?

Ultimately, though, such merger concerns tend to obscure what was really going on between the two companies—and more broadly, how power and control are actually exercised in modern capitalist economies. It’s unlikely Nadella and Microsoft ever actually intended to acquire OpenAI or its assets, or to directly hire Altman or his staff. The key point here is that Microsoft did not need to own OpenAI in order to control it, since it already benefited from its dominance over the company through exclusive licenses to technology and its power over OpenAI’s access to computing infrastructure. What would Microsoft gain from the added formality of bringing OpenAI or Altman’s team in-house?

What’s more, Microsoft potentially stood to lose a great deal in a move to own and administer OpenAI outright. It’s much easier to hold corporations responsible for actions they undertake through their own employees and business assets than via the arm’s length relationships with independent contractors and firms. As supply chains grow more complex through the heavy reliance on subcontractors, corporate chieftains operate within a far greater range of market impunity. In this case of OpenAI, Microsoft was able to exploit its hold on essential infrastructure and contracts blocking OpenAI from working with Microsoft’s rivals in a push for greater control over OpenAI, without having to face the many of the legal and regulatory consequences of a formal takeover. And by getting Altman his job back along with a new board, Microsoft expanded its control, while operating under with a liability shield that insulates it from much of the potential trouble OpenAI could get up to in the future.

Much recent antitrust commentary, including from top antitrust enforcers Jonathan Kanter and Lina Khan has focused on the need to rein in so-called vertical mergers. These are mergers at different levels of the supply chain, like Amazon’s purchase of movie studio MGM, or Microsoft’s acquisition of gaming company Activision Blizzard. In reality, though, vertical integration comes in many forms.

Consider Amazon, which outsources local delivery to nominally independent trucking companies operating exclusively for Amazon, with the retail monopoly dictating their routes, rates, and even pay rates. Why should Amazon go to all the trouble of vertically integrating through traditional ownership structures, when it can control drivers just as easily through these contracts, and in the process avoid a direct relationship with unionizing employees or liability for road accidents? In the same vein, Amazon uses various restrictive contracts to control the pricing policies of sellers on its platform and coerce them into using Amazon’s in-house fulfillment service. The logic here is identical: Why vertically integrate and acquire sellers, when you can simply control their businesses through contracts and collect a hefty fee on each transaction instead?

Such developments reflect long-brewing changes in market structures throughout the country. It’s true that, at certain points in the supply chain, horizontal concentration has escalated, as direct competitors have merged with each other, lessening the intensity of head-to-head competition. At the same time, however, vertical integration has actually been declining in importance as a mechanism of corporate control. As sociologist Gerald F. Davis has pointed out, traditional measures of corporate size like total assets, revenues, market capitalization, and employment that had been tightly correlated with each other in the past are now being unwound. Large corporations, as measured by revenue are often not large employers anymore—they instead disburse liability and accountability through subcontracted supply chains.

Take the airline industry. Many large airlines have bought up their direct competitors, reducing the number of nationally competitive airlines to just four (American, United, Delta, and Southwest). But amid this concentration, air carriers have also cut back on vertical integration, by spinning off and outsourcing ticketing, maintenance repairs, and other key functions to lower-wage subcontractors. On feeder routes to their hub airports, airlines have even outsourced the flying of the airplanes themselves to regional airlines—which, despite what the paint on the plane and the uniforms of the flight attendants might suggest, are legally separate airlines.

This shift from vertical integration to vertical outsourcing happened as the result of deliberate market positioning. To begin with, corporations lobbied heavily to outsource economic activity to subcontractors, independent contractors, franchisees, and other legally separate companies by substituting restrictive contracts—known in antitrust jargon as “vertical restraints”—for formal ownership. These vertical restraints formerly ran afoul of antitrust law—particularly when bigger firms imposed them on less-powerful independent businesses. But a battery of key court decisions in the 1970s loosened previous restrictions on these contracts. So, just as unions and other observers warned, corporations began to maintain tight control over their supply chains and business networks, while also carrying out an outsourcing binge that effaced many of the obligations, legal duties, and headaches that had previously accompanied formal vertical integration. This restructured model of corporate operations was especially effective in blocking drives to organize unions: It was far easier to relinquish a contract with a unionizing contractor than to crack down on one’s own unionizing employees—as Amazon has just done to one of its unionizing delivery contractors in Palmdale, Calif.

Vertical outsourcing also gained dramatic new momentum thanks to changes in information and communications technologies. All these measures made it much easier for corporations to prescribe, monitor, and control the activities of “independent” subcontractors. The vertically integrated corporations of an earlier era had to rely on telephones, fax machines, and in-person supervision—but now major employers like McDonald’s have nearly real-time access to on-the-job performance of cashiers at nominally independent franchised restaurants. Similarly, GPS technology allows Amazon to dictate delivery drivers’ routes and know at all times the precise locations of its delivery fleet, even while the company’s outsourcing contracts disclaim responsibility for accidents or the Taylorite conditions contracted drivers face.

Microsoft had already achieved a similar measure of arm’s-length control over OpenAI—notably via the provisions in its deal with the company forbidding its managers from taking its technology to a rival cloud computing partner. Microsoft never had any need or desire to take the extra step of formally acquiring a majority stake in OpenAI and appointing its own board of directors.

Microsoft’s subsequent personnel putsch drives home a crucial point for the next stage of antitrust enforcement. Going forward, we need to reconsider how we regulate business practices that allow large corporations to gain control over economic activity while sidestepping fundamental accountability for this increased market power. That means, for starters, rethinking the basic ways we frame and enforce bans on vertically integrated mergers. Indeed, in many of these contracted arrangements, vertical mergers would be preferable. Vertical contracts too often allow corporations to avoid responsibilities for activity they control—but vertical integration, whatever its other ills, lacks this particular vice. If Amazon were to directly own its trucking fleets, workers would not have to struggle as hard as they do now to hold it accountable for their working conditions. In fact, Microsoft itself just demonstrated this benefit of vertical integration: It recently agreed with the Communications Workers of America union to stop subcontracting certain subcontracted workers and bring them on as direct Microsoft employees, netting immediate wage and benefit gains for unionized workers.

An alternative approach to antitrust in the vertical outsourcing sphere should start with a new battery of curbs on contract-driven corporate integration. We also must strengthen labor and employment law, together with vicarious liability laws, tax rules, and other bodies of law and regulation. These revisions should flow from one core principle: It should not be so easy to disclaim employment responsibility by hiring through a temp agency, or to escape accident liability by outsourcing to a third party.

We should also rein in the ability of corporations to control third parties through contract in the first place. If McDonald’s executives want to maintain tighter control over local restaurant operations, prices, and staffing, they should take responsibility if a customer gets sick, or if workers’ wages are stolen. With the machinations in and around OpenAI firmly in mind, we must bring an end to the age of corporate impunity.