In a gig economy built on information asymmetry, DoorDash drivers in the United States accepted food delivery assignments without knowing the tip amount in advance. The platform withheld the figure until after the driver committed to the trip. A driver had no mechanism to distinguish a two-dollar job from a twenty-dollar one before accepting. A small company called Para identified the asymmetry and built a corrective: its application intercepted the data packet sent to the driver’s phone, extracted the concealed tip, and displayed it. Para gave workers information they were practically and arguably contractually entitled to receive. DoorDash shut Para down within months — not through superior engineering, but through a law passed in 1998. The story is not an edge case in the history of the digital economy. It is the central case. The enshittification documented in Part 1 of this series did not emerge from unconstrained market forces operating on neutral terrain. It was constructed through three specific policy failures. And it is sustained, above all, by a legal architecture that criminalizes escape.
Key Takeaways#
The consumer welfare standard, adopted in the 1980s, replaced structural antitrust enforcement with a harm-to-prices test that is structurally incapable of identifying harm in zero-price markets.
Facebook’s 2012 acquisition of Instagram — one employee per approximately $77 million paid — was permitted under a review framework that required demonstrated consumer harm rather than structural market analysis.
Regulatory capture in the tech sector operates through geographic arbitrage: routing European operations through Ireland’s under-resourced data protection authority neutralized GDPR enforcement against the largest platforms.
Between 2022 and 2024, approximately 500,000 technology sector job eliminations removed individual exit as a disciplinary constraint on internal product decisions.
Section 1201 of the Digital Millennium Copyright Act makes it a federal felony to circumvent a digital lock, converting the platform’s preferred information asymmetry into a legally enforceable condition.
The corrective is legal and institutional: interoperability mandates, portability rights, and the decriminalization of circumvention tools are the structural equivalents of the competitive markets that the consumer welfare standard eliminated.
The App That Knew Too Much#
Para’s shutdown illustrates a dynamic that repeats across the digital economy with consistent structure. A platform constructs an information asymmetry that benefits the platform at the expense of the party with less structural power. A third party builds a tool to correct the asymmetry. The platform uses intellectual property law to eliminate the tool. The asymmetry is restored not by market competition — a competing platform offering greater transparency — but by legal enforcement of the incumbent’s preferred architecture. This is not the market functioning. It is the market being suspended by statute.
The three constraints that once made this kind of platform behavior commercially unsustainable — competitive pressure that would redirect users to less extractive alternatives, regulatory enforcement capable of penalizing the behavior, and worker power capable of refusing to build the systems that enable it — have been methodically removed over four decades through policy choices that are documented, attributable, and reversible. Understanding the current condition of the digital economy requires understanding how each constraint was dismantled and what specifically was used to dismantle it.
Three Missing Constraints#
The platform economy’s capacity for continuous extraction rests on the structural absence of three institutional checks that disciplined corporate behavior through most of the twentieth century. Each was eroded through specific, identifiable policy choices rather than through the natural operation of technological change. Each created conditions that made enshittification not merely possible but economically rational. Reconstructing the conditions under which extraction becomes commercially punishable does not require new technology. It requires reversing the policy decisions that removed the original constraints.
How the Walls Were Built#
Competition as a Policy Relic#
The Sherman Antitrust Act of 1890 rested on a premise articulated clearly in its congressional debates: concentrated commercial control over the necessities of daily life is structurally analogous to political autocracy and equally incompatible with the conditions required for a functioning democratic order. The legislation gave federal authorities a mandate to prevent dominant market positions from becoming self-perpetuating. For roughly a century, the enforcement was imperfect and uneven, but the structural logic remained operative. Dominant firms faced at least the credible possibility of antitrust action, which constrained the most aggressive consolidation strategies.
The displacement of this framework began in the 1980s with the adoption of the consumer welfare standard, associated primarily with the legal scholar Robert Bork. The standard held that market concentration was economically benign provided it did not produce demonstrably higher prices for consumers. A firm achieving a ninety percent market share was, under this reasoning, demonstrating superior efficiency by revealed preference — the market had selected it. The standard created a logical circularity from which enforcement could rarely escape: dominance was evidence of quality, and quality was evidence of legitimacy. Preventing a firm from becoming dominant was tantamount to penalizing success.
The standard was structurally incapable of addressing harm in zero-price markets. If Google’s search product is nominally free, there is no consumer price to measure harm against. If Facebook’s social platform costs nothing in money, the traditional harm analysis produces no result. The consumer welfare standard was not designed for markets where the medium of extraction is attention and data rather than price. Applied to those markets, it functioned as a structural exemption from antitrust scrutiny for the most significant commercial consolidation of the twenty-first century.
The practical consequences are documented. Facebook’s 2012 acquisition of Instagram for approximately one billion dollars — a company with thirteen employees — was approved without structural intervention. Internal Facebook communications produced in subsequent litigation confirmed that Zuckerberg’s purpose was preemptive elimination of a competitive threat rather than integration of complementary capabilities. By 2022, independent analysts estimated Instagram’s standalone value at over one hundred billion dollars. The merger that converted a potential competitive alternative into a consolidated property of the incumbent was permitted because the consumer welfare standard provided no mechanism to anticipate harm in a market where the price to users was zero. The structural precondition for enshittification — the absence of a credible exit option — was constructed with regulatory consent.
The Regulator in the Revolving Door#
When competitive markets fail to discipline corporate behavior, the theoretical corrective is regulatory enforcement. The European Union’s General Data Protection Regulation, enacted in 2018, represented the most ambitious attempt to construct such enforcement in the digital economy. It established explicit rights of access, erasure, and opt-in consent for personal data processing, with penalties of up to four percent of global annual revenue for violations. Applied to the largest technology platforms, those penalties would represent sums in the billions of dollars. The GDPR appeared to impose a structural cost on the surveillance-based extraction model that Part 1 of this series documented.
American technology companies responded with a structural countermove rather than compliance. They concentrated their European operational headquarters in Ireland, a decision whose logic was transparent. Ireland competes for multinational corporate presence through a package of institutional accommodations: corporate tax rates among the lowest in the OECD, a permissive regulatory climate, and an English-speaking workforce. The Irish Data Protection Commission inherited supervisory jurisdiction over the European operations of Google, Meta, Apple, and Microsoft simultaneously. Between GDPR’s enactment in 2018 and 2022, the Irish authority issued fewer major enforcement decisions against these companies than any comparable national supervisory body in the EU. The penalties it did impose were a fraction of the statutory maximum and insufficient to alter the economics of the behavior being penalized. A law is only as effective as its enforcer. The enforcer had been selected through the mechanics of regulatory arbitrage.
The lobbying infrastructure that sustains this environment reinforces the dynamic. Between 2017 and 2023, the five largest American technology companies spent a combined total exceeding five hundred million dollars on federal lobbying in the United States. This expenditure is not anomalous corporate behavior. It is the predictable output of oligopolistic market structure. A market of a hundred competing firms cannot coordinate a unified lobbying strategy — the firms compete against each other for regulatory advantage as well as for customers. Five dominant firms share a concentrated structural interest: preventing regulatory intervention that would alter market architecture. Consolidation converts the regulatory environment from a site of competition into a site of coordination. The same concentration that produces enshittification produces the political capacity to sustain it.
When the Engineer Could No Longer Say No#
For much of Silicon Valley’s formative decades, software engineers occupied a structural position of unusual individual power. A skilled engineer at a major technology company contributed, by internal estimates cited in congressional testimony and industry surveys, approximately one million dollars annually to the firm’s operating margin. Industry growth consistently exceeded the rate at which the labor market could supply qualified engineers. An engineer who objected to an ethically compromised product decision could resign and receive an equivalent offer within a week. This exit option, exercised by even a small proportion of the workforce, functioned as an informal but operative constraint. The capacity to build a system that defrauded users depended on the willingness of engineers to build it. If they refused at sufficient scale, the system did not get built.
Between 2022 and 2024, the technology sector eliminated approximately five hundred thousand jobs across Google, Meta, Microsoft, Amazon, and dozens of smaller firms in a pattern that appeared coordinated in its timing and scale. The surplus of available engineers that resulted removed the credible exit option that had functioned as an internal check. The constraint did not fade gradually. It was removed through a deliberate cycle of hiring expansion followed by abrupt contraction that returned leverage to management.
The legal architecture that completes this structure was constructed in 1998 through the Digital Millennium Copyright Act. Section 1201 of the legislation established a federal criminal offense — carrying penalties of five years imprisonment and a five hundred thousand dollar fine — for circumventing a digital access control mechanism. The European Parliament enacted equivalent provisions in 2001 through Article 6 of the Copyright Directive. Both laws were framed as anti-piracy measures directed at unauthorized reproduction of music, film, and software.
Their operational scope proved far broader. A digital lock is now embedded in agricultural machinery, automobile computers, medical devices, printer consumables, and mobile applications. A farmer who accesses their own tractor’s diagnostic data using non-manufacturer software is circumventing a digital lock. An independent mechanic who resets an automobile’s engine management system after a repair that does not use manufacturer-certified parts is circumventing a digital lock. A developer who builds an application that intercepts and displays information a platform would prefer to withhold — as Para did with DoorDash’s tip data — is circumventing a digital lock. All face criminal exposure under the statute whose stated purpose was preventing music piracy. The effect of Section 1201 in the platform economy is to make the incumbent’s preferred information asymmetry legally enforceable. Extraction that would be commercially defeated by a competitive corrective tool is instead sustained by the threat of prosecution.
This is why fifty-one percent of internet users have installed ad-blocking software on web browsers while zero percent have installed equivalent tools inside mobile applications. Web browsers operate on open standards. Applications are closed architectures wrapped in intellectual property protection. The browser’s openness is what makes the corrective tool legal. The application’s closure is what makes the equivalent tool a federal crime.
The Reversible Condition#
The enshittification of the digital economy is not a technological inevitability produced by the inherent nature of network effects or platform economics. It is a legal construction, assembled from identifiable components: the consumer welfare standard that exempted zero-price markets from antitrust scrutiny, the merger review regime that permitted the acquisition of competitive threats rather than requiring competition against them, the geographic arbitrage that neutralized GDPR enforcement, the labor market contraction that removed internal resistance, and the DMCA’s Section 1201 that converted exit tools into criminal instruments.
The implication is precise. The required interventions are legal and institutional. Better product design will not solve the problem — improved products will be deployed by the same firms under the same structural incentives, and improvements will be twiddled away when they conflict with extraction objectives. Privacy-conscious consumer behavior will not solve the problem — individual choices cannot overcome collective action failures at platform scale. The intervention identified by Doctorow and supported by a growing body of antitrust scholarship is interoperability: the legal right to build tools that connect to, extract data from, and enable departure from platforms without criminal penalty. Making it legal to display hidden gig-economy compensation data, to port a social graph between platforms, to use third-party diagnostic software on purchased agricultural equipment — these are not technological innovations. They are the structural equivalents of the competitive markets that the consumer welfare standard eliminated.
The historical precedent is clear. When the Bell System monopoly was dissolved in 1984 and telephone number portability was subsequently mandated, the switching cost that had sustained AT&T’s dominance dropped toward zero. Competitive entry immediately followed. The same structural logic applies to digital infrastructure. A platform that cannot retain users through exit costs must retain them through service quality. Enshittification is, in this precise sense, a direct function of the legally enforced impossibility of departure.
The internet’s foundational layer was built on open protocols. Email, web pages, and file transfer operate on technical standards that no single company controls or can monetize. The period of enshittification began when closed applications displaced open protocols as the primary interface of digital social and economic life. Reversing it requires no new engineering. It requires restoring the legal conditions under which the open-protocol model was viable: the right to interoperate, the right to exit, and the right to compete without requiring a billion-dollar acquisition to do so.
