“Trust Me Bro” Isn’t Going to Work for AI or Data Centers
For a very short time, I worked for Con Edison as a customer service representative (my foray into working exclusively from home, not for me, but I digress). That experience has stayed with me in ways I probably didn't fully appreciate at the time. When you work in customer service for a utility company, you get a very different understanding of what energy consumption means to an actual household. You learn about kilowatt usage, meters, billing cycles and energy efficiency, but you also learn what happens when someone's energy costs increase and there isn't more money in that household to absorb the increase. The calls that affected me most were from seniors, particularly people living on fixed incomes. Someone would call because their bill was higher than expected, and you could hear the concern because that wasn't just an inconvenient bill. That bill was competing with groceries, medication, housing costs and everything else that had to come out of a fixed amount of money every month.
In those situations, I would do what I could within the parameters of my job. I would go through the energy-efficiency recommendations we had available. I could look at usage and try to determine whether something seemed unusual. In some cases, I could arrange for a technician to check the meter. You try to cross the t's and dot the i's because you understand that the person on the other end of the phone isn't calling to have a theoretical conversation about electricity consumption. They are trying to figure out whether they can afford to continue living the way they have been living. That experience is part of the reason I think we need to be much more careful about the way we are discussing artificial intelligence, data centers and the enormous amount of energy infrastructure that is going to be required to support them.
Technology companies tend to discuss energy in gigawatts. Investors discuss capital expenditures. Policymakers talk about generation capacity, transmission and economic competitiveness. Those are all legitimate ways to understand what is happening, but that's not how an ordinary person experiences energy systems. People experience energy through their utility bills. They experience infrastructure through what gets built in their communities, what happens to their roads, what happens to their water supply, what happens to their taxes, what happens to their electricity rates and whether the economic development happening around them creates an opportunity that they can actually access.
Think about someone who has worked for decades, paid off a home, retired and intentionally gotten their cost of living to a level that they believe their retirement income can support. Social Security cost-of-living adjustments already have to stretch across food, housing, healthcare and everything else becoming more expensive. Now imagine that person hearing that a technology company worth hundreds of billions of dollars wants to build an enormous AI data center in or near their community, and one of the first things they hear about that facility is how much electricity and water it could require. Their first reaction isn't excitement about American AI leadership. It's, "What is this going to do to my bill?" This is a reasonable question and it's one I think anyone serious about building the infrastructure necessary for the next era of American tech leadership should be prepared to answer…specifically.
This is why I believe America's AI infrastructure challenge is becoming something bigger than an infrastructure challenge. It is becoming a trust challenge. The industry needs enormous amounts of electricity, land, water, transmission capacity, skilled labor and political cooperation to build what it believes will be necessary over the next decade. But it also needs something that doesn't appear on a data-center construction budget: public trust! If people increasingly believe that the AI transformation means technology companies receive the economic upside while communities receive higher costs, greater resource constraints and very little say in the process, the industry should expect resistance. That resistance can't be dismissed as people being afraid of technology or failing to understand innovation. The industry needs to start reckoning with why people are becoming skeptical in the first place.
AI and Data Center’s “Trust Me Bro” Problem
A couple of months ago, I wrote on LinkedIn that tech had a narrative problem around artificial intelligence and that if the industry didn't start correcting it, the issue could become a bomb by the midterm elections. What I was concerned about then wasn't whether people liked AI. My concern was that the technology industry was allowing a much broader story to develop about who benefits from technological advancement, who bears its costs and whether ordinary people have any recognizable place in the economic future that Silicon Valley keeps describing (“describing” - we’ll come back to that). Now, as the political conversation around AI data centers accelerates, I think we are watching that narrative problem become physical.
The fundamental mistake tech is making isn't a communications mistake. It's not that the industry needs a better advertising campaign explaining why artificial intelligence is good. It is that the distributed benefits of AI aren't recognizable enough to ordinary people, and there is no sufficiently clear timeline for when those benefits will become recognizable. Meanwhile, the concerns people associate with the transformation feel immediate. People are watching layoffs now. They are worried about job security now. They feel the pressure of inflation and higher living costs now. They are hearing about increased electricity bills now. They are hearing about water consumption demand now. They are watching technology companies spend extraordinary amounts of money on AI infrastructure now. Then the industry comes along and says that AI is going to increase productivity, create new industries, increase GDP and create new kinds of opportunities… AT SOME POINT.
The problem is that, to someone who is already disgruntled, that sounds like “Trust me bro.”
Trust me bro, the opportunities are coming.
Trust me bro, productivity growth will eventually improve living standards.
Trust me bro, the jobs that disappear will be replaced by new jobs.
Trust me bro, this enormous infrastructure buildout will be good for your community.
Trust me bro, the economic benefits will eventually spread outward from the companies and investors making enormous amounts of money today.
This sounds like "AI Reaganomics," which we know didn't work.
I happen to believe there are very real reasons to be optimistic about what AI could contribute to American economic growth and productivity. But optimism about the future doesn't answer someone's concerns in the here and now. You can't respond to present-day anxiety with a futuristic narrative and then be surprised when people don't want to hear it.
Social media makes this problem even more difficult for the tech industry because people can see dissatisfaction developing in communities they would never otherwise hear about. Someone in Pennsylvania can see a video about a data-center fight in Virginia. Someone in Ohio can watch a resident somewhere else talk about water consumption. Someone in New Jersey can see a local government meeting from Arizona. People can compare utility bills, job experiences, layoffs, zoning disputes and political spending in real time. Whether every claim circulating on TikTok, X, Reddit, Facebook or YouTube is accurate is a separate issue. From a trust perspective, what matters is that people can see other people expressing the same frustration they are feeling, and that creates a feedback loop that a corporate communications strategy can't easily control.
That is why Chamath Palihapitiya's warning about data centers becoming a symbolic representation of asymmetric upside for a narrow tech elite resonates with this moment. His framing gets at something deeper than electricity consumption. Data centers are enormous, expensive and physically visible representations of an AI economy whose benefits many people still believe are concentrated among a relatively small number of companies, executives, investors and highly compensated tech workers. The problem isn't that the public has performed a detailed economic analysis and concluded that the benefits will never spread. The problem is that the industry hasn't shown them convincingly enough how, when and through what mechanisms those benefits reach them.
I don’t feel that data centers created tech's trust problem. However, data centers gave the trust problem a physical address. AI used to be largely abstract to most people. It was ChatGPT on a laptop, an AI-generated image on social media, a chatbot at work or a conversation about whether a machine could eventually replace some part of someone's job. A data center changes the relationship because now AI requires something from a physical community. It requires land. It requires electricity. It requires water. It requires substations, transmission infrastructure, construction workers, permits, tax arrangements and political decisions. The industry is no longer asking people to adopt a product. It is asking communities to host the physical infrastructure of an economic transformation.
That requires strengthening the (already eroding depending on who you ask) social contract!
We Have Enough Frameworks
The numbers around data-center electricity consumption make it clear why this conversation is ratcheting up. Different studies produce different projections depending on assumptions and methodology, so I don't think we should treat any single forecast as the end all. What is much harder to dispute is the direction of travel. Bluefield Research has projected that data centers could account for approximately 8.9 percent of U.S. electricity demand by 2030. Other projections put the potential share higher (as high as 17%). Axios has reported on updated Berkeley Lab estimates that could put data-center electricity consumption somewhere between roughly 9.5 and 15.3 percent by 2030. Longer-term Bloomberg projections have contemplated data centers consuming as much as approximately one-fifth of U.S. electricity by 2035. These are projections, not guarantees, but even the lower ranges imply an extraordinary infrastructure challenge if AI investment continues anywhere close to its current trajectory.
The water issue also demonstrates why this conversation has to be approached through systems thinking rather than isolated talking points. Bluefield Research has projected that by 2030 a substantial majority of the total water consumption associated with data centers could occur off-site and be tied to electricity generation rather than onsite cooling. That matters because the public conversation separates electricity and water into two different boxes. In reality, the systems interact. Electricity generation can require water. New generation can require new transmission. New transmission requires construction. Construction requires workers. Workers require training. New industrial loads require utility investment. Utility investment raises questions about who pays. Who pays raises questions about ratepayers. Ratepayers vote. Voters influence politicians. Politicians influence permitting, regulation and the speed at which infrastructure can be built. There is no serious way to discuss AI infrastructure while pretending these are separate, isolated issues.
This brings me to Pennsylvania Governor Josh Shapiro and his GRID standards. There is a lot about GRID that I think moves the conversation in the right direction. Pennsylvania's approach addresses energy affordability, transparency and community engagement, workforce and economic development, and environmental protection. Developers seeking Commonwealth support are expected to provide information about projected electricity demand and water consumption, develop community outreach plans, address workforce development and explain how infrastructure costs associated with their projects will be handled.
Shapiro is also widely discussed as a potential 2028 presidential candidate, and I think it is fair to ask whether his decision to become increasingly visible and aggressive on the data-center issue may have something to do with the political environment he expects to encounter if he pursues a national campaign. Obviously I don't know his motivations on this data center issue, but frankly, if a politician who may want to run for president is looking at the public backlash around data centers and concluding that this is an issue voters care about, I am fine with that. That's politics. People express concern, politicians realize those concerns matter electorally and policy begins to respond.
My issue is what comes next. We have enough frameworks. I am not saying GRID is meaningless. In fact, it is more substantive than many frameworks because it attempts to attach actual requirements to state support. But the word “framework” has become one of those words institutions use when they want to demonstrate that they understand a problem without necessarily demonstrating how an ordinary person will experience the solution and that’s only Step 1. I want Step 2 through Step 5. What happens to someone's electricity bill? What protections exist for a retired homeowner living on a fixed income? Who pays for a new substation? What infrastructure does the utility need? What jobs become available locally? What training begins before the jobs arrive? Where can residents see the projections? Who represents residents when infrastructure agreements are being negotiated?
That's the level of detail required if the objective is trust rather than policy compliance. “Community engagement” can't mean residents are invited to a public meeting after developers, lawyers, consultants, economic-development officials and government agencies have spent months discussing a project.
Protect the Ratepayer, Then Build the Workforce
This is why Congressman Byron Donalds's Protecting Ratepayers Act is the type of initiative I want to see policymakers exploring. Donalds's proposal starts with a principle people can understand immediately: ordinary households shouldn't be forced to absorb infrastructure costs created by massive private data-center developments. His legislation seeks to require data centers to independently source water and power rather than straining public infrastructure, codifying a ratepayer-protection principle into law. Whatever one thinks about the precise mechanics of the legislation, the political concept is clear enough to explain at someone's kitchen table. A technology company making billions of dollars should not be able to build an enormously energy-intensive facility and simply pass the infrastructure burden down to the retired homeowner living three miles away.
That is an important start, but protection from downside can't be the entire social contract. If a community is going to host infrastructure that is supposedly essential to America's economic future, people should be able to see how they can participate in the upside. This is where I think the conversation around workforce development has been astonishingly incomplete. We talk constantly about preparing people for an AI workforce, but that usually means teaching people how to use AI tools, code, work with data or understand machine learning. Those things matter too, but AI also has a physical workforce. Somebody has to build the electrical infrastructure. Somebody has to maintain the grid. Somebody has to work on transmission lines. Somebody has to build and maintain substations. Somebody has to install industrial electrical systems. Somebody has to maintain cooling systems. Somebody has to repair the equipment when it fails.
The Bureau of Labor Statistics projects electrician employment to grow 9 percent from 2024 to 2034, with approximately 81,000 openings per year on average over the decade. For electrical power-line installers and repairers, BLS projects approximately 10,700 openings annually. For electrical and electronics installers and repairers, approximately 9,600 annual openings are projected, much of that reflecting workers leaving occupations or retiring. These jobs aren't all being created by AI, and we shouldn't pretend they are. The workforce projections existed within a much broader infrastructure and energy environment. But if we are simultaneously forecasting significant new electricity demand, grid modernization requirements and enormous private investment in AI infrastructure, then we should be asking whether our education and workforce systems are preparing people for the physical economy supporting that investment.
This is where I am very serious about Career and Technical Education. If America's electrical infrastructure is aging (and our infrastructure definitely isn't keeping up with China's), if electricity demand is expected to grow, and if technology companies are planning data-center investments on a scale that could reshape regional power systems, then electrical education should become a national CTE priority. I want to see that reflected in federal funding asap. For example, by 2028, every federally funded CTE system in America should be required to demonstrate that interested students have an accessible pathway into the electrical-energy workforce. That does not mean every individual high school has to build an identical electrician program. Different regions have different employers, licensing systems, facilities and needs. A pathway might involve a regional technical school, community-college dual enrollment, a utility partnership, registered apprenticeship, union training or work-based learning. But there should be a pathway, and locally specific pathways at that.
And I want the definition of that pathway to be broad enough to reflect the infrastructure we actually need. Electrical trades. Power-line work. Substations. Industrial electrical systems. Energy systems. HVAC and cooling. Grid technology. Electrical engineering technology. Utility careers. These aren't secondary to the AI economy. They are part of the physical foundation underneath it. If we can forecast how much compute capacity we think America will need in 2030 and 2035, we can forecast the human capacity required to support it. This is a great place for the American public to see distributed benefits of AI and data centers now.
This is where we can start correcting the “Trust me bro” problem in a tangible way. Imagine a 16-year-old student in Pennsylvania whose community is debating whether to approve a new data center. Instead of that student hearing only that AI might eventually create jobs, imagine that the local CTE school, utility company, community college and developer can show that student an actual pathway. Here are the courses you take. Here are the credentials. Here is what the work pays. Here are the employers. Here is how you can move from technician to another role later. Here is how you could eventually start your own electrical contracting business. Now the economic benefit isn't an abstract promise about GDP in 2035. There is a door standing in front of that young person today.
If we are going to tell people that the next twenty years require enormous infrastructure investment, we should simultaneously tell young people how they can participate in building it. There are many examples in American history when we moved rapidly national economic and security priorities required workforce mobilization. There is no reason workforce development should trail behind capital deployment now. We shouldn't wait until utilities, contractors and developers are screaming about shortages before deciding that maybe we should have started training people five years earlier.
You Can’t Run Roughshod Over the Communities You Require Buy-in From
There is another part of this conversation that tech and the political organizations surrounding it need to understand. People are paying attention to how money moves through politics. They are paying attention to lobbying. They are paying attention to Super PACs. They are paying attention when a community raises objections through the channels government tells them to use and then watches enormous amounts of outside money influence the political process. Whether every perception is fair isn't the issue, but the perception itself affects legitimacy.
Tech money can't run roughshod over communities and then expect those communities to embrace technological transformation. That approach reinforces the exact narrative the industry needs to dismantle: a small group of extraordinarily wealthy companies and tech bros receive the upside, while everyone else gets told what is going to happen. You can't fix that perception by spending even more money telling people they misunderstand what is good for them. Again, "Trust me bro."
This is particularly important because winning a formal decision isn't the same thing as winning legitimacy. A company can win a zoning vote and lose the community. It can obtain a permit and deepen distrust. A political organization can win an election and make the underlying resentment worse. Systems have consequences beyond the immediate transaction, and I think the tech industry sometimes approaches political problems too transactionally. The objective becomes getting the permit, getting the tax treatment, getting the land, winning the vote or defeating the opposition. But if each victory produces another layer of public distrust, eventually every future project becomes more difficult.
There are practical ways to do this differently. If a utility says a region requires major grid investment to accommodate projected demand, let the public hear the utility leaders explain it. Put the conversation online. Stream it through local government websites. Let residents hear what the utility needs from the developer, what it needs from the state government and what it believes it needs from the federal government. If water consumption is a concern, explain the assumptions in language people can understand. Explain direct consumption. Explain indirect consumption. Explain what happens during drought conditions. Explain what conservation requirements exist. If jobs are being used to justify a project, publish the (proposed/forecasted) jobs. Separate temporary construction employment from permanent employment. Publish salary ranges and required credentials. Identify the training partners early, ahead of time. Identify apprenticeships. Establish local hiring goals where appropriate. Then return a year later and tell the public whether those commitments were met (my suggestion is via video, again hosted on local government websites).
Libraries should be part of this conversation. CTE schools should be part of it. Community colleges should be part of it. Senior centers should be part of it. Utility leadership should be part of it. Local government should be part of it. We keep saying that artificial intelligence will transform society and then designing the public conversation as if the only people who need to understand the transformation are engineers, investors and policymakers. If this is societal transformation, then society needs mechanisms for understanding and participating in it.
This is also why I included the recent vandalism of Flock license-plate cameras in my thinking about data-center resistance. A Flock camera and an AI data center are obviously not the same thing. The privacy and surveillance concerns associated with automated license-plate readers are different from the electricity, water, land-use and economic questions associated with data centers. I'm also not predicting that people are going to begin destroying data centers. Let's be clear, this isn't my claim. The comparison is about what can happen when technological disagreement turns into resistance because people believe systems have been imposed on them without meaningful consent.
The Washington Post's reporting on people destroying Flock cameras should be understood as a warning about tech legitimacy. When people conclude that the formal mechanisms available to them aren't working, resistance can move outside those mechanisms. That should matter enormously to an industry preparing to spend extraordinary amounts of money on physical infrastructure. The capex involved in AI data centers is entirely too high for tech companies, investors and Super PACs to think they can run roughshod over everyday citizens, get the infrastructure built and deal with public anger later.
The smarter approach is partnership before resistance hardens. Give people information. Give them meaningful participation. Give them protections. Give them economic pathways. Let them see how decisions are made. Let them see what utilities are saying. Let them understand what government officials are negotiating. Let them know what companies have promised. Then create mechanisms through which those promises can be tracked.
Public Trust Is Infrastructure
Maybe what we need is something like an AI Infrastructure Community Impact Statement. Before a major AI infrastructure project receives final public support, communities should have access to one understandable place where they can see what is actually being proposed. Not a 700-page filing written primarily for lawyers, engineers and consultants. Something a resident can reasonably read.
Who is building the project? Who will use the facility? How much land does it require? What is its expected peak electricity demand? Where will that electricity come from? What generation, transmission or distribution infrastructure must be built? Who pays for it? What is the estimated direct water consumption? What relevant indirect water impacts exist? What tax incentives are being provided? How many construction jobs are expected? How many permanent jobs? What are the salary ranges? What credentials will workers need? Which local institutions will provide training? What apprenticeships will exist? What local hiring commitments have been made? What environmental safeguards apply? What happens if the developer fails to meet its commitments? Who monitors compliance? Where does the public go five years later to see whether the promises became reality?
Parts of Shapiro's GRID standards already move in this direction, which is why I think the next stage should be about turning these ideas into something residents can use. Make the information accessible. Make it comparable. Make it replicable across projects. Make it something a retiree can understand, a high school guidance counselor can use, a local journalist can track and an elected official can pull up five years later and ask, “You promised this. Did you do it?” That is how frameworks become governance.
No single actor can accomplish this. Tech companies can't fix the trust problem alone (or cram acceptance down people's throats). Utilities can't fix it alone. Governors can't fix it alone. Congress can't fix it alone. Schools can't fix it alone, and communities can't fix it alone. That's why this is a systems problem. Tech companies need to become more transparent about the physical requirements of their AI ambitions and more concrete about the opportunities those investments create. Utilities need to explain what the grid can support, what needs to be built, how long it takes and who should bear the costs. State and local governments need to protect residents while establishing predictable rules for responsible development. The federal government needs to stop treating AI policy, energy policy and workforce policy as siloed conversations. CTE schools and community colleges need to start building workforce capacity before shortages become crises. Libraries and community institutions need resources that allow ordinary people to understand what is happening without first becoming electrical engineers.
All of them need to get on this now, because trust takes much longer to build than a data center does. Once distrust becomes embedded, every subsequent decision becomes more difficult. Every zoning application looks suspicious. Every utility proposal looks suspicious. Every promise of jobs looks suspicious. Every claim about national competitiveness looks suspicious. At some point, the industry can provide all the data in the world and discover that people no longer trust the institutions providing the data.
I believe America needs to remain a tech leader. I believe artificial intelligence can create enormous economic value. I believe we need to continue to build. I believe we need more energy infrastructure, more grid capacity and the compute required to compete in a world where AI increasingly influences economic and geopolitical power. None of those beliefs requires me to dismiss the concerns of the people being asked to live alongside the infrastructure. There is a version of America's AI future in which we build enormous amounts of compute, create extraordinary wealth, lead the world technologically and still leave millions of people believing the transformation happened around them rather than with them. That could look like tech success on a spreadsheet or earnings call while being a profound systems failure.
There is another version in which the AI infrastructure boom becomes the catalyst for grid modernization, electrical workforce development, CTE pathways, apprenticeships, community investment, utility transparency and a much more sophisticated relationship between tech companies and the communities hosting their infrastructure. That is a different social contract, and I believe we still have time to build it.
Data centers are becoming symbols. Politicians are noticing. Communities are organizing. Elections are approaching. Technology companies are committing capital at a scale that makes the consequences of getting this wrong increasingly expensive. So maybe the most important infrastructure question facing AI and data centers right now isn't how many gigawatts America can build. Maybe it’s, what would America's AI buildout look like if public trust were treated as seriously as compute capacity? What would change if workforce development started before the data center arrived? What would change if a parent could look at a proposed project and identify real career pathways for their teenager? What would change if a retired homeowner could see exactly how their utility bill would be protected? What would change if communities could see electricity assumptions, water assumptions, tax incentives, employment commitments and infrastructure costs before the consequential decisions were made? What would change if community participation stopped being treated as an obstacle to innovation and started being treated as part of the infrastructure required to sustain it?
Because the answer can't continue to be “Trust me bro.”
Show people the jobs. Show them the training. Show them the protections. Show them the numbers. Show them the infrastructure requirements. Show them who pays. Show them what their community receives. Give them a meaningful voice before the decisions have already been made, and then keep showing them what happens after the ribbon cutting. Let me repeat, I said, "SHOW," not "TELL."
If AI really is going to transform American economic life, ordinary Americans can't merely be consumers of that transformation. We have textbooks. Crack 'em open and draw on lessons learned from WWII.
We can't build the infrastructure of the future while treating public trust as an afterthought.
Sources
Business Insider — Chamath Palihapitiya / AI data-center backlash:https://www.businessinsider.com/chamath-palihapitiya-ai-data-center-backlash-warning-2026-8
Bloomberg. “AI Data Centers Are Sending Power Bills Soaring.” Bloomberg, September 2025 - https://www.bloomberg.com/graphics/2025-ai-data-centers-electricity-prices/
Bluefield Research — data-center electricity and water demand:https://www.bluefieldresearch.com/ns/data-center-electricity-demand-nearly-doubles-to-8-9-by-2030/
Axios — AI power/data-center electricity and construction: https://www.axios.com/2026/08/19/ai-power-data-center-electricity-construction
Pennsylvania Governor’s Office — GRID standards: https://www.pa.gov/governor/newsroom/2026-press-releases/gov-shapiro-releases-full-grid-standards-to-protect-pennsylvania
Pennsylvania Governor’s Office — legislation to codify GRID standards: https://www.pa.gov/governor/newsroom/2026-press-releases/news--pa-house-passes-legislation-to-codify-gov-shapiro-s-grid-s
Axios — Josh Shapiro / data-center policy shift: https://www.axios.com/2026/08/19/josh-shapiro-ai-data-centers-pivot
Axios — Shapiro and possible 2028 positioning: https://www.axios.com/local/philadelphia/2026/08/20/gov-josh-shapiro-s-latino-ad-blitz-is-an-early-handshake-with-voters-he-may-need-in-2028
Rep. Byron Donalds — Protecting Ratepayers Act: https://donalds.house.gov/news/documentsingle.aspx?DocumentID=2635
U.S. Bureau of Labor Statistics — Electricians: https://www.bls.gov/ooh/construction-and-extraction/electricians.htm
U.S. Bureau of Labor Statistics — Line installers and repairers: https://www.bls.gov/ooh/installation-maintenance-and-repair/line-installers-and-repairers.htm
U.S. Bureau of Labor Statistics — Electrical and electronics installers and repairers:https://www.bls.gov/ooh/installation-maintenance-and-repair/electrical-and-electronics-installers-and-repairers.htm
U.S. Bureau of Labor Statistics — Electrical and electronics engineers:https://www.bls.gov/ooh/architecture-and-engineering/electrical-and-electronics-engineers.htm
U.S. Bureau of Labor Statistics — HVAC mechanics and installers: https://www.bls.gov/ooh/installation-maintenance-and-repair/heating-air-conditioning-and-refrigeration-mechanics-and-installers.htm
Washington Post — Flock camera vandalism / surveillance backlash: https://www.washingtonpost.com/nation/2026/08/18/vandals-are-destroying-license-plate-cameras-amid-anger-over-surveillance/
Related Reading
EPRI: https://www.epri.com/research/sectors/technology/results/3002034696
Fast Company: https://www.fastcompany.com/91581106/can-ai-solve-the-energy-problem-it-created-ai-data-centers-technology-energy
Wall Street Journal: https://www.wsj.com/tech/inside-big-techs-frantic-race-to-quell-the-growing-backlash-to-ai-2a717339
