DEV Community

Keith MacKay
Keith MacKay

Posted on

The Companies Disrupting Your Job Are Now Proposing Rules to Protect It

#ai #career #management

The Companies Disrupting Your Job Are Now Proposing Rules to Protect It

The robot tax, the four-day workweek, the public wealth fund. Good ideas, self-serving ideas, and missing ideas. Let's sort them out.

You might have noticed that the companies building the technology most likely to displace workers are now publishing policy papers about how to protect workers. OpenAI released "Industrial Policy for the Intelligence Age" in April 2026: a 13-page document addressed primarily to US policymakers, proposing robot taxes, public wealth funds, a shorter workweek, and a new social contract for the AI era [1].

Is the paper worth your time? Some of it is genuinely forward-thinking. Some of it is a "policymercial": advocacy dressed as policy analysis, and a way for the industry leaders to try to shape the policy conversation that (as always) is slower than the evolution of the technology. As a result there are some policy ideas that are not covered at all.

What Happens If We Do Nothing

Before sorting through proposals from vested stakeholders, it's worth thinking about what the stakes are. The paper is not catastrophizing. The risks it names are real and already in motion.

The labor market shock is not hypothetical. Anthropic's labor market research found measurable displacement effects concentrated in software development, customer support, and data analysis: precisely the knowledge worker categories that expanded most dramatically over the past two decades [2]. Workers in the most exposed roles are 47% higher-paid and more educated than average: the people who least expected to be displaced. Young workers aged 22-25 in exposed occupations are already finding jobs 14% less often than they were before ChatGPT's release. Aggregate unemployment statistics don't show it yet. The leading edge does.

The energy grid is already straining. Data center electricity consumption could reach 1,050 TWh globally by 2026, placing the sector among the largest power consumers on the planet [3]. In Ohio, electricity prices rose from 11-12 cents per kilowatt hour in 2020 to 19 cents in 2025 [4]. Dominion Energy proposed its first base rate increase since 1992, adding roughly $8.51 per month to a typical household's bill [4]. The cost of AI infrastructure doesn't disappear when a company pays its cloud bill. It distributes to ratepayers who didn't sign up for it.

The wealth concentration problem compounds. As OpenAI's own paper acknowledges: as AI replaces labor, wealth accrues to owners of capital rather than owners of labor. Tax systems built on payroll taxes and labor income lose their base. Social Security, Medicaid, and housing assistance all depend on labor income as their funding mechanism. If labor income shrinks and capital income grows, the safety net becomes structurally underfunded exactly when demand for it rises.

The numbers are no longer theoretical. In 2026, Block cut 4,000 workers (40% of its workforce) with CEO Jack Dorsey explicitly attributing the cuts to AI and agentic workflows [16]. Meta, Intuit, and Atlassian followed with cuts of 10%, 17%, and 10% of their workforces respectively [17]. Tech layoffs reached 142,000 in 2026, with roughly half attributed to AI automation [18]. Worth noting: UVA's Darden School questioned whether Block's framing was "AI strategy or AI scapegoat" (Dorsey also admitted to over-hiring during COVID), but the trend across companies is too consistent to explain away.

The disruption is compounding in a second way. A May 2026 Gartner study of 350 global executives found that organizations cutting workers to demonstrate AI returns are not seeing them. Companies using AI to amplify workers outperformed automation-only strategies. Gartner VP Helen Poitevin: "Workforce reductions may create budget room, but they do not create return." [19] The companies dismantling their workforces are likely making a strategic mistake as well as a social one.

None of this requires a dystopian scenario. It only requires the trends already underway to continue at their current trajectory.

The Experts Disagree. Their Incentives Don't.

The gap between the most optimistic and most pessimistic economic forecasts for AI may be the largest disagreement in modern macroeconomics.

Daron Acemoglu, the MIT economist who shared the 2024 Nobel Prize in Economics, ran the numbers. His paper "The Simple Macroeconomics of AI" uses Hulten's theorem: GDP impact equals the fraction of tasks affected times the average cost savings. His conclusion: AI will add roughly 0.07% to annual TFP growth over the next decade (about 0.71% in total) [5]. In a May 2026 interview, Acemoglu identified three signals to watch: whether AI agents can handle the multi-task fluidity actual jobs require (he's skeptical), whether the wave of AI apps comparable to transformative software like Excel ever materializes (still missing), and whether the accelerating race by AI companies to hire prominent economists ends up producing genuine research or just expensive hype management [6].

On that last point, he was direct: "What I hope we won't get is that they're interested in economists just to further their viewpoints or further the hype." OpenAI has recruited Jason Furman (former Obama economic adviser) and installed Ronnie Chatterji as its first chief economist. Anthropic has assembled a 10-member economic advisory council. Google DeepMind hired Alex Imas as "director of AGI economics." Every major AI lab now employs senior economists whose incentives run directly toward findings that support their employers' continued growth.

Dario Amodei, CEO of Anthropic, is at the opposite end of the forecast range. At Davos in January 2026, he said: "I can see a world where AI brings the developed world GDP growth to something like 10 or 15 percent." He added that the scenario could produce "5% to 10% GDP growth together with an unemployment rate of 10%," acknowledging that "that's not a combination we've almost ever seen before" [7]. The gap between Acemoglu's 0.07% annual and Amodei's 15% GDP scenario represents roughly $100 trillion in cumulative global output over a decade. It is, straightforwardly, a bet on whether AI will be incrementally useful or historically transformative. (Amodei is building a company valued at over $60 billion on the transformative premise. His forecast is also his fundraising pitch.)

Jamie Dimon occupies the pragmatic middle. In May 2026, JPMorgan's CEO said the bank will "hire more AI people and fewer bankers in certain categories." He warned that the pace of AI adoption "may go too fast for society" and gave a specific example: two million commercial truckers displaced too quickly could spark civil unrest [8]. His prescription is collaborative public-private management of the transition. His concern, notably, is not purely altruistic. Rapid mass unemployment destabilizes the financial system. Social unrest is bad for banks. JPMorgan has roughly 25,000 to 30,000 departing employees per year through natural attrition; Dimon can afford to absorb AI displacement gradually precisely because of that turnover buffer. His call for a managed transition is also risk management.

Gregory Daco, Chief Economist at EY-Parthenon, offers the most grounded near-term read. The data already shows what he calls "a clear decoupling between growth and hiring": output expanding while companies generate that growth with fewer workers and fewer hours [9]. But he's cautious about the causal story. Only about 10% of firms currently use AI to produce goods and services. He notes he's "not entirely sure this is a replacement situation where talent is being replaced by technology." At least not yet at scale.

Anthropic's own Economic Index provides the most specific ground-truth data available. Only 4% of jobs use AI for 75% or more of their tasks. About 36% have some AI involvement for at least a quarter of their work [2]. The gap between theoretical capability and actual deployment remains wide. The signal most worth watching: a 14% reduction in job-finding rates for workers aged 22 to 25 in highly AI-exposed occupations since ChatGPT's release [2]. Aggregate unemployment statistics don't yet show it. The leading edge does.

Paul Krugman captures the distributional challenge in a sentence: "How do we figure out a system for not just having prosperity, but for having shared prosperity?" [10] He and MIT's Erik Brynjolfsson both flag a J-curve dynamic: productivity gains lag the initial investment as companies retrain workers and redesign processes. The gains, if Amodei is even partially right, could be extraordinary. Whether they are broadly shared is not an economic question. It's a policy one.

The Proposals (and What's Behind Each One)

The Robot Tax

OpenAI proposes shifting the tax base from labor income and payroll taxes toward corporate income and capital gains, including potential levies on AI-driven automation [1]. The logic: if a company replaces ten workers with AI, it no longer pays its share of the payroll taxes those workers generated. Taxing the automation compensates for the lost base.

This has precedent. South Korea reduced tax incentives for companies investing in automation in 2017, effectively raising their relative tax burden compared to human employment. The EU has debated robot taxes repeatedly. Bill Gates proposed one explicitly in a 2017 interview, arguing that taxing robots at rates equivalent to displaced workers would fund retraining and social investment [11].

MIT economists modeling the optimal robot tax rate suggest a range of 1-3.7%: high enough to recapture displaced payroll tax revenue without driving automation offshore or pushing companies toward human-capital-light business models that simply move jobs rather than create them [20]. The design of the rate matters as much as the principle.

The idea also benefits OpenAI directly, though the paper doesn't say so plainly. A robot tax creates legitimacy for AI adoption by giving governments a fiscal mechanism to offset disruption, reducing political pressure to slow or block AI deployment. The tax pays for the permission.

The Public Wealth Fund

Modeled on Norway's sovereign wealth fund and Alaska's Permanent Fund, the proposal calls for a government-owned fund investing in AI companies and AI-adopting firms, distributing returns to citizens [1]. The logic: if you can't work your way into the upside of AI, at least own a piece of it.

Alaska's Permanent Fund has distributed $1,000-$2,000 annually to every resident since 1982, funded by oil revenues [12]. Norway's Government Pension Fund Global has accumulated over $1.7 trillion [13]. The mechanisms work. The design question is: who seeds the fund, and with what?

In June 2026, Senator Bernie Sanders announced he would introduce the American AI Sovereign Wealth Fund Act: a one-time 50% equity tax on OpenAI, Anthropic, and other large AI companies, seeding a government-owned fund to distribute direct cash payments to Americans [21]. Sanders explicitly cited OpenAI's own April 2026 paper as precedent. The proposal is blunter than what OpenAI had in mind. The company proposed that "policymakers and AI companies work together" on fund design. Congress heard "50% equity tax" and ran with the logic. When you propose a public wealth fund without specifying who pays for it, someone else will specify it for you.

OpenAI leaves that deliberately vague. "Policymakers and AI companies should work together" is not a funding plan. A fund seeded with equity from frontier AI firms (including OpenAI) would spread the upside. A fund seeded with tax revenues from existing businesses would just be a new layer of redistribution. The difference matters enormously.

The self-interest here: if citizens have financial stakes in AI success, they become politically invested in AI's continued growth. Public opposition to AI development becomes harder to sustain when opposition means opposing your own quarterly dividend.

The Four-Day Workweek

OpenAI proposes that productivity gains from AI be converted into shorter working hours: a four-day or 32-hour workweek with no reduction in pay [1]. This has actual experimental evidence behind it. Iceland ran the largest public sector four-day workweek trial in history from 2015 to 2019; researchers found maintained or improved productivity and measurably better worker wellbeing [14]. Subsequent trials across hundreds of companies in the UK, Ireland, and Australia have shown similar results.

The appeal is obvious: if AI makes each worker more productive, sharing that productivity gain as time rather than headcount reduction keeps employment levels stable and improves worker wellbeing. The alternative is layoffs, which concentrates the gain at the ownership level.

This one benefits OpenAI somewhat indirectly: it reduces political hostility to AI and makes the case that AI can be net-positive for workers. But it's also the proposal with the strongest independent evidence base and the least obvious hidden angle.

The Expanded Safety Net

The paper recommends boosting retirement contributions, expanding healthcare coverage, and subsidizing child and elder care for workers in AI-affected sectors [1]. This is largely a restatement of longstanding progressive labor policy wearing new clothes. It's not wrong. Worker transitions are easier when the safety net is thicker. The self-interest here is again about political legitimacy: AI companies benefit from a policy environment where disruption is manageable, not destabilizing.

Energy and Infrastructure

The paper calls for public-private partnerships to accelerate grid expansion, with a specific note that "households should not be subsidizing AI data centers" [1]. This is a striking thing for OpenAI to publish, given that current utility rate structures are already pushing data center costs onto residential ratepayers.

More than 30 states are now proposing or implementing tariffs that require large-load customers (data centers) to pay higher rates to cover grid infrastructure costs [4]. The argument is straightforward: if you're adding load to the grid at scale, you should pay for the infrastructure that load requires, not spread it across residential users who consume a fraction of what you do.

The self-interest is obvious: if data centers pay their own infrastructure costs, cloud economics stay competitive. If residential ratepayer backlash gets bad enough to trigger regulatory restrictions on data center siting and power access, that's a direct constraint on AI infrastructure expansion. Getting ahead of this with a principled-sounding position is smart positioning.

What's Missing

The OpenAI paper is notably light on ideas that would benefit society at the cost of AI companies. Critics at TechPolicy.Press specifically noted the paper contains no mention of antitrust enforcement, data privacy protections, structural power asymmetries between AI companies and users, or geopolitical competition: omissions that collectively leave the paper unable to address the market concentration risks that may matter most [22]. Here are several proposals that have come from other quarters:

Stronger liability for AI-displaced workers. Some labor economists have proposed requiring companies to pay severance and retraining costs proportional to the number of AI-displaced roles, creating a direct financial obligation that internalizes the cost of displacement rather than socializing it. This would genuinely make AI adoption more expensive and slow the pace of workforce disruption.

AI deployment moratoria in specific sectors. Proposals from labor unions and some European policymakers include mandatory impact assessment periods before AI deployment in high-employment sectors, with worker consultation requirements. France and Germany have both floated versions of this for public sector employment. Slowing deployment is not in OpenAI's interest.

Mandatory worker ownership stakes. Several Nordic labor models require significant worker ownership in companies above certain sizes. Applied to AI companies, this would give labor a direct claim on AI-generated productivity rather than relying on government redistribution. OpenAI's paper talks about workers sharing in gains; it doesn't propose worker ownership mechanisms.

Data rights and compensation. The training data that makes large models capable came overwhelmingly from human creative and intellectual labor (writers, coders, artists) who received no compensation. Multiple legislative proposals in the EU and several US states would require licensing payments or data use fees for training data. This would materially increase AI development costs and constrain what models can be trained on. The paper doesn't address it.

Open-source AI mandates. Regulatory proposals from some researchers and public interest advocates would require frontier AI models above certain capability thresholds to release weights publicly after a defined period, cutting the ability of a small number of companies to capture long-term returns from AI. This would benefit the public while reducing the moat of companies like OpenAI.

Energy quotas for AI data centers. Direct caps on data center electricity consumption, or carbon pricing mechanisms that make compute more expensive, would slow AI infrastructure expansion and reduce grid stress. Several European countries are already considering this. The paper argues for better grid infrastructure rather than consumption limits: a position that happens to align precisely with AI companies' growth plans.

Independent economic research, free from AI company funding. Acemoglu's concern about AI labs hiring economists to shape narrative deserves its own policy response. Public funding for truly independent AI economic research, at arm's length from companies with trillion-dollar stakes in the conclusions, is one of the simplest and most overlooked proposals. It does not appear anywhere in the OpenAI paper.

Worth noting: Anthropic published its own competing policy paper, "Preparing for AI's Economic Impact," in early 2026 [15]. It is somewhat more candid than OpenAI's. Anthropic explicitly acknowledges that compute and token taxes would "directly impact Anthropic's revenue" and includes them anyway as options for severe disruption scenarios. That transparency about self-interest is rare in this genre, and the Anthropic paper is worth reading alongside the OpenAI one.

Sorting the List

The proposals in the OpenAI paper range from genuinely good ideas with real evidence (the four-day workweek), to smart market-legitimizing moves dressed as altruism (the robot tax), to vague gestures pending real design work (the public wealth fund). The missing proposals (liability, moratoria, worker ownership, data rights) share one characteristic: they would cost AI companies something.

That doesn't make the OpenAI paper cynical. Companies advocate for policy environments that allow them to operate. That's normal. The paper is more thoughtful than most corporate policy advocacy, and the risks it identifies are accurate.

But a complete picture of the policy landscape requires reading it alongside the proposals its authors chose not to make. The full toolkit available to society for managing AI disruption is much wider than what any AI company will voluntarily recommend.

The Bottom Line

The expert forecast range runs from 0.07% annual TFP growth to 15% GDP, and everyone with an opinion has skin in the game. The risks of doing nothing are real: already visible in electricity bills, in slowing hiring for young workers in exposed occupations, and in Dimon's warning that the pace of change "may go too fast for society." OpenAI's proposed solutions are a mix of good ideas, market-protecting moves, and incomplete designs. The solutions absent from the paper share one trait: they would cost AI companies something.

The companies building the technology most likely to displace workers are also now building the economic research teams, the policy papers, and the narrative about what solutions should look like. Read the OpenAI paper. Read Anthropic's too. Then read Acemoglu. Then ask whose fingerprints are on the analysis you're using to make decisions.

What proposals do you think deserve more attention in the policy debate around AI and the economy? Which of the OpenAI proposals do you think would actually get traction?

References

  1. Industrial Policy for the Intelligence Age, OpenAI, April 2026
  2. Labor market impacts of AI: A new measure and early evidence, Anthropic
  3. Energy demand from AI, IEA
  4. AI data centers use a lot of electricity. How it could affect your power bill, NPR
  5. The Simple Macroeconomics of AI, Daron Acemoglu, NBER Working Paper 32487, 2024
  6. Three things in AI to watch, according to a Nobel-winning economist, MIT Technology Review, May 2026
  7. Anthropic CEO Dario Amodei warns of 5-10% GDP growth with 10% joblessness, Yahoo Finance / Davos, January 2026
  8. Dimon: AI's effect on labor market 'may go too fast for society', Banking Dive; see also JPMorgan will hire more AI specialists, fewer bankers, Bloomberg, May 2026
  9. Productivity gains fuel U.S. growth while hiring slows, Gregory Daco / Yahoo Finance
  10. How should we think about the economics of AI?, Paul Krugman / Substack (featuring Erik Brynjolfsson)
  11. Bill Gates: The robot that takes your job should pay taxes, Quartz, 2017
  12. Alaska Permanent Fund Dividend, Alaska Department of Revenue
  13. The fund's market value, Norges Bank Investment Management
  14. Going Public: Iceland's Journey to a Shorter Working Week, Autonomy / ALDA, 2021 (summary of the 2015-2019 government trials)
  15. Preparing for AI's economic impact: exploring policy responses, Anthropic, 2026
  16. Block, Jack Dorsey's company, cuts 4,000 workers — CEO says AI is taking their jobs, Fortune, February 2026
  17. Companies replacing workers with AI in 2026, Tech.co
  18. Tech layoffs reach 142,000 in 2026 as profitable companies cut jobs to fund $700B AI infrastructure, TechTimes, May 2026
  19. Gartner Says Autonomous Business and Artificial Intelligence Layoffs May Create Budget Room But Do Not Deliver Returns, Gartner, May 2026
  20. Universal Basic Income and Automation: The Optimal Robot Tax, First Movers AI
  21. Sen. Bernie Sanders teases bill to give public ownership of AI companies, Washington Times, June 2026
  22. The Doublespeak in OpenAI's Industrial Policy for the Intelligence Age, TechPolicy.Press

If this resonated, here are some related articles:

Keith MacKay is a technology strategy consultant and CTO in EY-Parthenon's Software Strategy Group (SSG), specializing in AI disruption and technology diligence for private equity and corporate clients. SSG's AI Disruption Lab conducts rapid assessments of how AI transforms and threatens existing business models and value chains. Keith teaches at Northeastern University and writes about strategy, management, and AI/technology, with Claude Code and Codex as AI collaborators.

Top comments (0)