Fast Company
Ý tưởng kinh doanh mới từ Fast Company
-
Knowledge workers judge AI governance
For years, brand reputation was built on straightforward outputs: the product worked, the service delivered, and the mission statement said the right things. But, behind the scenes, a sophisticated cohort has quietly moved the goalposts.
Data from our latest 2026 Brand Expectations Index reveals that knowledge workers—the professionals closest to enterprise buying decisions, talent pipelines, and industry conversations—have changed how they evaluate a company. They no longer just audit what your company does. They are evaluating how your company decides when to use AI.
As autonomous systems scale, this audience is looking past slick interfaces to inspect the governance behind the product, the judgment behind the claim, and the explicit human accountability behind the system. For marketing and communications teams, this structural shift changes the rules of the game.
Comfort is not an endorsement
It is easy for brands to mistake market familiarity for actual trust. Knowledge workers are highly comfortable with artificial intelligence as an operational layer, especially when compared to the general public. Our research shows a high baseline level of comfort with companies deploying AI for marketing (77%), personalization (78%), and customer service workflows (76%).
However, knowledge workers’ comfort hits a hard ceiling the moment AI moves from routine automation into autonomous, decision-making roles:
- 65% are comfortable with AI automating critical security functions.
- 58% resist AI making HR decisions.
- 55% reject the technology generating legal or policy documents.
This isn’t a contradiction; it’s a clear market signal. Knowledge workers have separated two questions that most brands still mistakenly treat as one: Is AI useful? and Should AI be deciding this? They have answered an emphatic yes to the first. The answer to the second depends entirely on the transparency of your governance.
Brands that communicate under the assumption that product adoption equals cultural endorsement are completely misreading the room.
Capability claims aren’t enough
The current corporate communications landscape is overcrowded with capability-driven messaging. Companies rush to announce what their AI models can do, how fast they operate, and the efficiency gains they unlock.
Fewer explain what AI should not do, where human review explicitly intervenes, and who ultimately owns the outcome when a system fails. For a highly discerning audience, those narrative gaps don’t read as corporate nuance. They read as operational risk.
According to our index data, 63% of knowledge workers want to see companies consult outside experts before deploying higher-stakes AI initiatives. Furthermore, 66% rank a leader’s long-term reputation—defined by demonstrated judgment over time, rather than media visibility or category hype—as a primary driver of trust.
This audience isn’t looking for a flawless corporate record; they operate inside complex organizations and understand technical trade-offs. What they demand is verifiable evidence that a human being remains fully accountable for the machine’s choices.
Context over volume
To build real trust in an AI-driven market, communications leaders must lead with the reasoning, not just the result. When announcing an AI deployment, your narrative must proactively answer the three questions your buyers are already asking internally:
- Why did you deploy it here?
- Where do the guardrails live?
- Who owns the fallout?
The leaders successfully building premium brands are explicit about where the software ends and where human oversight begins.
Our data suggests that audiences heavily reward this operational context. Last year, our study found that 84% of knowledge workers rank direct communications from companies—long-form articles, executive platforms, and transparent owned content—as a top-tier trusted source of information, second only to local news. They don’t want a higher volume of content; they want a higher caliber of context.
The trust gap
This demand for rigorous corporate decision-making has created a massive, overlooked opening for emerging companies.
While only 28% of the general population trusts AI startups, that number more than doubles to 58% among knowledge workers. This massive trust gap represents an extraordinary strategic window. Right now, however, most AI startups are burning that advantage by defaulting to generic category language, inflated claims, and use cases that read more like fleeting tech demos than durable enterprise value.
The precise audience most likely to champion your adoption inside the enterprise is also the cohort most sensitive to corporate overclaiming. They can instantly hear the difference between an AI company that has done the actual work on governance and one that is merely performing it.
Knowledge workers are not a forgiving audience, but they are highly receptive to brands that have earned their position. They are not looking for leadership teams that project absolute certainty in an uncertain market. They are looking for organizations that demonstrate consistent, verifiable, and rigorous judgment in what they build, how they deploy it, and how honestly they communicate about both.
The next definitive test of AI market leadership will not be a question of who moves fastest. It will be a question of who can make the judgment behind the technology visible and worthy of trust.
Tyler Perry is co-CEO at Mission North.
-
Workers are interested in creative AI tools. But hardly any are using them
Artificial intelligence is changing how people view creativity in the workplace. While some leaders have touted the technology as improving the creative process at their companies, there are broader concerns about outsourcing the creative process to AI.
This week, an Adobe survey revealed that 90% of workers want to use creative AI tools at work, but only 9% actually report implementing any in their workflow. Adobe defined creative AI as tools that are applied to visual or imaginative tasks, like image and video generation, designing graphics, brainstorming ideas, and creative writing.
The self-reported survey pulled from 1,002 full-time U.S. employees who use AI at work at least once a week, across industries like finance, healthcare, retail, education, and more.
According to the survey results, there seems to be a strong interest among employees in creative AI tools, but daily use does not necessarily reflect that enthusiasm in real life. Instead, respondents said they mainly use AI on technical tasks, like summarizing documents, analyzing data, and researching topics.
Adobe also found that 72% of workers surveyed believe AI could potentially make them more creative at work. For those who already do use creative AI tools, 73% said it improves their ability to relay complex ideas.
Still, employees estimated they spend an average of 7.5 hours per week on tasks that could potentially be enhanced by creative AI tools. And another one in three workers said they would like to use AI for image and video generation, but don’t.
There could be a few reasons for the disconnect, according to the survey respondents.
Nearly half of those surveyed (47%) said they feel unsure about how creative AI can apply to their job. Some 29% reported a lack of training or resources and unclear company policies as contributors to their lack of use of the tools, while others noted difficulty integrating AI into their workflows (27%) and the cost of AI tools (26%) as factors.
AI adoption also varies by industry. Those employed in the tech and innovation sector are most likely to say their company is ahead of the adoption curve, while those working in education reported feeling like the least likely to be ahead.
Other studies have document how employees are using AI at work.
In a recent Gallup poll, 47% of U.S. employees said their organization has adopted AI tools to improve productivity, efficiency, or quality—up from 41% in the previous quarter. More than half of U.S. workers polled (52%) said they use AI in their role, with 30% using it a few times a week or more, and 15% using it daily. The most common AI uses reported are writing and editing (51%), search or research (49%), and general assistance or problem-solving (39%).
Just 17% of AI users polled by Gallup reported using it for image, video, or audio generation.
The biggest gap between employee enthusiasm and real-world execution may have to do with the infrastructure around it. Employees responding to the Adobe survey revealed a few action items that could help them increase their use of creative AI—like being offered examples of how it can be used in their roles, better integration, employer-provided training, and more assurance about data privacy.
Still, companies that do embed AI tools in tasks like brainstorming ideas, creative writing, or image generation will likely have to weigh any creativity gains with the risk of overreliance—and not replacing human critical thinking skills that drive innovation.
-
Move to one of these 25 ‘cities on the rise’ if you want better job opportunities in 2026
LinkedIn has released its second annual “Cities on the Rise” list, ranking the top 25 U.S. metro areas that it says have emerged as the fastest-growing destinations for jobs and new talent.
The Microsoft-owned social network analyzed its labor market data to determine where professionals are landing jobs, employers are hiring, and new graduates are launching their careers.
Augusta, Georgia, earned the top spot for 2026. Other cities on the list include Harrisburg, Pennsylvania; Louisville, Kentucky; Richmond, Virginia; Little Rock, Arkansas; and Tulsa, Oklahoma.
For those exploring new career opportunities, this list offers insight into high-growth cities for jobs and new talent. Check out the full list below.
Top 25 cities for 2026:
- Augusta, Georgia
- Richmond, Virginia
- Reno, Nevada
- North Port-Bradenton-Sarasota, Florida
- Harrisburg, Pennsylvania
- Charleston, South Carolina
- Pensacola, Florida
- Austin, Texas
- Portland, Maine
- Tulsa, Oklahoma
- Lancaster, Pennsylvania
- Albany, New York
- Portland, Oregon
- Indianapolis, Indiana
- Charlotte, North Carolina
- Myrtle Beach, South Carolina
- Appleton-Oshkosh-Neenah, Wisconsin (Fox Cities)
- Fort Wayne, Indiana
- Sacramento, California
- Little Rock, Arkansas
- Salt Lake City, Utah
- Wilmington, North Carolina
- Louisville, Kentucky
- Milwaukee, Wisconsin
- Grand Rapids, Michigan
Here’s why Augusta is ranked No. 1
Augusta sits at the top of the list. If you’re looking for new career opportunities and have the flexibility to relocate, you may want to consider looking for work here, LinkedIn says.
Top industries hiring in the region include higher education, hospitals and healthcare, and electric power generation. The city’s top employers include Savannah River Nuclear Solutions, the U.S. Army, and Augusta University.
The U.S. Army Cyber Command at Fort Eisenhower is based here, as is the Georgia Cyber Innovation & Training Center, which is the country’s largest state-owned cybersecurity facility. Beyond defense and cybersecurity, growing industries include medtech and data centers.
The median household income is $55,485, while the average home listing price is $396,195. LinkedIn data shows that from February 1, 2024, to February 1, 2026, 26.5% of advertised jobs in the metro area were remote and 3.5% were hybrid.
More than half of the cities were featured in last year’s list
A year ago, LinkedIn released its inaugural “Cities on the Rise” list. Augusta didn’t earn a spot on last year’s roundup.
However, 60% of the cities from that ranking returned this year, including Richmond, Virginia; Harrisburg, Pennsylvania; Albany, New York; Myrtle Beach, South Carolina; and Milwaukee, Wisconsin.
Grand Rapids, Michigan, which held the top spot in last year’s ranking, fell to 25th place this year—a dramatic shift. Forty percent of the U.S. metros highlighted this year are making their first appearance.
New metros include:
- Augusta, Georgia
- North Port-Bradenton-Sarasota, Florida
- Charleston, South Carolina
- Tulsa, Oklahoma
- Lancaster, Pennsylvania
- Charlotte, North Carolina
- Appleton-Oshkosh-Neenah, Wisconsin (Fox Cities)
- Little Rock, Arkansas
- Salt Lake City, Utah
- Louisville, Kentucky
-
A fake Temu-like website claims to give shoppers a dopamine boost—without buying anything
Shoes by ThreadTheory. Snacks by CrunchCove. Gadgets by SkyScribe. They’re all for sale on FakeHaul.com, a new competitor to online marketplaces like Temu. The site offers the same bottom-of-the-barrel prices and massive selection of products. The only difference? None of the brands, products, or prices on FakeHaul are real—and that’s entirely by design.
FakeHaul is meant to replace sites like Temu not by offering better products or lower prices, but by offering no products at all. Instead, it aims to replicate the experience of online retail, presumably scratching the same itch as a shopping spree without costing users a cent.
The rise of dopamine sites
FakeHaul is the latest arrival in the “dopamine site” category, an emerging genre of entertainment-style apps that strive to give users the same dopamine release as online shopping, sans the credit card bill at the end of the month. You can fill up your cart, make customizations, hunt for deals, and even check out, all without actually purchasing any products or providing personal information like a shipping address or a payment method.
The site was created by Jack Yang, a software developer working on AI at Microsoft. FakeHaul was a passion project for Yang, inspired by someone in his life who struggles with impulse shopping.
“I noticed the satisfaction mostly happened before the purchase: in the browsing, the comparing, the adding to cart, and the waiting for delivery,” Yang tells Fast Company over email. “I wanted to see if that ritual could be separated from the financial consequence.”
Later in development, Yang took cues from FoodNeverComes, a similar so-called dopamine site that mimics food delivery apps like DoorDash. There’s also Dopamine Shop, a shameless imitation of Amazon, and No Smoke Zone, a cigarette simulator that aims to curb cravings for nicotine rather than online shopping.
“The dopamine-site stuff helped me realize this was a broader behavior and gave me language for it, but the personal use case came first,” Yang explains. “FoodNeverComes does it for a single fake delivery; FakeHaul runs the whole shopping loop.”
Though Yang first experimented with showing real products from brands like Gucci and Louis Vuitton on the site, he swapped them out for AI-generated substitutes to avoid legal trouble when FakeHaul launched. In its first two days of being live, the site garnered 550 page views from 100 visitors—a small group, but one that showed genuine interest and follow-through on the site’s features.
“A good share of people who landed went deep into the funnel rather than bouncing off the joke, which is the behavior I was hoping for,” Yang says.
A solution, or a new problem?
While FakeHaul was designed to replace addictive shopping sites like Temu and Shein, it contains some habit-forming features of its own. There’s a once-a-day wheel spin to unlock a coupon, and a shareable receipt after completing a fake order, showing how much money a user saved by shopping on FakeHaul instead of a real online store.
Yang admits that he was conflicted about adding those features to FakeHaul. From his perspective, the receipt system “is meant to flip the usual reward” of online shopping, letting users show pride in how much they didn’t spend. “Making it shareable was an attempt to make restraint feel like the thing worth showing off,” he adds.
The coupon wheel is “the more ethically complicated piece,” Yang says. “Yes, it’s built to bring people back, but ideally as the thing you open when you feel the urge to open Temu or Shein instead. Capping it at once a day was a small guardrail, so the substitute doesn’t quietly become another endless loop of its own.”
Yang continues: “The gap between parodying an engagement loop and simply running one is thin, and I’m not completely sure which side I land on.”
Yang has already received mixed feedback on the efficacy of FakeHaul. The user he had in mind when he designed the site says it helps her. “She feels the urge, browses, and ‘buys’ on FakeHaul, and it passes without a real purchase,” Yang says.
But others struggling with compulsive online shopping feel the opposite. When Yang went to the moderator of a subreddit for shopping addicts and asked if he could share FakeHaul with its users, he was shut down. The moderator explained that “rehearsing the buying ritual can reinforce the loop that people in recovery are trying to break, and a fake hit might not be a clean substitute,” instead sending them back down the rabbit hole of shopping addiction.
“That’s why I keep calling this an entertainment product with an open question inside it and not a treatment. I don’t have enough users yet to say which view the evidence backs,” Yang says. “If an overwhelming [number] of people say it becomes a warm-up [rather] than a healing ritual, I have a kill switch to shut the site down because it was not my original intention.”
-
Is it safe to eat lettuce again? The answer is more complicated than you think
A diarrhea-causing illness has taken over the U.S. and the news cycle, but as officials try to trace its origins, many greens lovers are asking themselves the same question: Can I eat lettuce now?
The answer is maybe.
Cyclosporiasis has spread throughout much of the U.S., with the multistate outbreak already affecting at least 1,645 people and another 5,000 cases still under investigation. The major outbreak has put food under added scrutiny—particularly lettuce. But the specifics have turned increasingly confusing. Here’s what we know.
Last week, the Food and Drug Administration (FDA) linked the outbreak to iceberg lettuce from a Taylor Farms de Mexico facility in central Mexico, although the FDA later announced the tests were actually a false positive. Still, Taylor Farms has issued a voluntary recall of the product, and the FDA recommends avoiding products recalled by the company. The iceberg lettuce is the only identified source at the moment, although investigations in the U.S. and Mexico remain active.
According to the product recall announcement, the affected lettuce includes batches distributed from June 29 through July 16 across various states, with “best used by” dates running through August 3. The recall includes a variety of products that feature shredded, chopped, and salad-mix versions.
The affected lettuce has been distributed across the following 27 states: Alabama, Arkansas, Connecticut, Florida, Georgia, Illinois, Indiana, Iowa, Kansas, Kentucky, Louisiana, Maryland, Massachusetts, Michigan, Mississippi, Missouri, New Hampshire, New Jersey, North Carolina, Ohio, Oklahoma, Pennsylvania, South Carolina, Tennessee, Texas, Virginia, and Wisconsin.
What lettuce is affected?
It’s not only bags with the Taylor Farms name that may be affected. Some lettuce products from the Marketside brand come from the same facility and may be available at Walmart—the retailer has since removed products suspected to be affected as a precaution.
Additionally, many restaurants and school cafeterias receive their produce from one of the nation’s largest food suppliers, Sysco, which carries products from Taylor Farms. Notably, Taco Bell has also been impacted by the outbreak. Reportedly, Taco Bell has removed the recalled lettuce from its supply chain following links to the cases in Indiana, Kentucky, Michigan, Ohio, and West Virginia.
For those who believe they may have bought affected lettuce, they should compare the product code and date at the back of the packaging with the details shared on the Taylor Farms website.
Cyclosporiasis is a form of intestinal illness caused by consuming produce or water infected with fecal matter. Triggered by a microscopic parasite, the illness takes anywhere from days to weeks to manifest, leaving those infected with symptoms like bloating, cramps, and explosive diarrhea.
According to reports, the parasite is especially hard to test for due to other compounds in food and the low number of infectious eggs present on produce, forcing federal investigators to rely heavily on patient interviews to track the source.
Can I eat lettuce now?
As experts urge consumers to remain vigilant about the raw produce they consume while the outbreak is ongoing, these experts also point out that not all lettuce is affected by the recall.
Although past outbreaks affected other forms of produce like blueberries, cilantro, basil, snow peas, and other leafy greens, there is no direct link to the current outbreak and these items should be safe to eat for now. For consumers looking to take extra precautions, leafy greens and other fresh produce can still be consumed by cooking them at 158 degrees Fahrenheit or higher. -
Nobody knows how bad corporate AI emissions really are. This startup has a way to estimate them
More and more companies are implementing artificial intelligence into both their internal workflows and their external products, and that AI use comes with an environmental impact. Already, AI data centers are driving a surge in electricity demand that is outpacing supply.
But that impact likely isn’t showing up on all corporate sustainability reports yet, because accounting for corporate AI emissions is a challenge—particularly when companies are using closed AI models that don’t disclose their energy use.
Watershed, a startup that helps companies track their emissions, is working on this challenge.
The startup recently published a framework for how companies can estimate their emissions from AI. It takes into account the data center infrastructure, a functional unit of kilograms of CO2 per million AI tokens, and calculations based on the number of AI tokens a company uses.
“Companies are already tracking AI usage at the token level for cost management,” John Bistline, Watershed’s head of science, tells Fast Company via email. “The emissions math plugs into that same data. So this isn’t asking companies to build something entirely new. Cost and sustainability go hand in hand here.”
Investors, auditors, and regulators are asking about corporate AI emissions
When companies quantify their carbon footprints, they take into account not only direct emissions from their own energy use or products, but indirect emissions, like from the flights their employees take for business travel—or all the power needed to answer their workers’ AI queries. Those are called Scope 3 emissions.
In some cases, Scope 3 emissions disclosures are already required by law, like in California. The Greenhouse Gas Protocol, which sets corporate standards, is considering requirements around cloud and AI services.
And aside from those requirements, companies are already being asked about these numbers.
“Investors, auditors, and regulators are asking about AI emissions, and most companies don’t have a defensible way to answer,” Bistline says.
Corporate AI footprints are growing
AI may be a small part of most companies’ footprints currently. “But nobody expects that to stay the case for long,” Bistline adds. “The companies that build their measurement infrastructure now will be better prepared than those who wait.”
By accounting for AI emissions, corporations will also be able to take steps to reduce both the emissions and their operating costs.
“The [Watershed] framework reports electricity alongside emissions specifically, so that measurement connects to concrete reduction levers: which model you use, which region serves your query, how you structure your prompts,” Bistline says. “Even with data gaps, these are all things companies can control in how they deploy and use AI.”
AI models can vary widely when it comes to energy use—a reasoning AI model may use about 30 times more energy than a smaller model for the same task, according to Watershed. “Region” also matters because different parts of the power grid are powered by different energy sources, which changes their carbon intensity.
Why AI emissions are still an estimate
Watershed’s framework only estimates the emissions from AI use. That’s because there’s no real way to precisely measure these emissions yet.
“Many of the most widely used AI models are closed, meaning you can’t independently test their energy consumption the way researchers can with open models,” Bistline says.
“The only empirical, published energy figure for a closed frontier model is Google’s Gemini data from mid-2025, and even that is a single data point for one model at one moment in time,” he adds.
Figuring out AI use emissions is also complex because of the research and development that goes into training these models. AI companies may not want to disclose the figures needed to do such calculations, either.
Those figures—concerning total training emissions and total lifetime tokens served—are “commercially sensitive,” Bistline says.
But even if AI providers won’t share those details, Watershed hopes they’ll share the ratio of emissions per token. (Tokens themselves are often a vague unit of measurement, adding to the challenge.)
That leaves an estimate of emissions as the best answer. As AI providers share more information, Bistline says, those estimates will get more precise.
And as AI providers share that info, it may show that their AI infrastructure is actually more efficient than the estimates assumed. That sort of disclosure, then, helps AI companies demonstrate their own efficiency gains as well.
-
Social Security benefits 2026 update: ‘Misleading’ email from Commissioner Frank Bisignano under fire
If you are one of the 63 million Americans currently receiving Social Security benefits, you’ll want to read on. (Some 54 million retired workers and 9 million of their survivors and dependents currently get a monthly check.)
Before we get started, here are some facts to keep in mind: The Social Security Administration (SSA) is an independent agency of the federal government. Currently led by Commissioner Frank Bisignano, the original “proponents of SSA’s independence wanted to insulate it from everyday political, fiscal, and operational policy decisions of the government.” That’s according to the agency itself.
Now for the news: Senate Democrats say a July 2 Social Security email to millions of retirees from Bisignano falsely credits Republican lawmakers and President Trump’s One Big Beautiful Bill Act (OBBBA) for giving older Americans tax credits.
The group of Democrats—including Sen. Elizabeth Warren of Massachusetts, Sen. Ron Wyden of Oregon, Sen. Tammy Baldwin of Wisconsin, Sen. Sheldon Whitehouse of Rhode Island, and Sen. Ben Ray Luján of New Mexico—say the email contains “a politicized message” and “misleading information.”
“Thanks to President Trump, over 35 million American seniors received an average of $7,500 in relief this tax season,” the email reads, with the subject line, “Making Life More Affordable for America’s Seniors.”
“President Donald J. Trump and the Trump Administration are not only protecting Social Security, but we are providing meaningful and immediate relief to older Americans,” the email continues.
However, critics—such as MS NOW—say the president didn’t exactly eliminate Social Security benefit taxes.
Fast Company has reached out to the Social Security Administration for comment.
What we do know is that Social Security is facing insolvency. As Fast Company has previously reported, the program’s retirement trust fund is estimated to run out of money in about seven years. One proposed solution is to limit payouts at $100,000 a year for couples as part of an overall plan to save it from insolvency, according to the Center on Budget and Policy Priorities (CBPP). (That amounts to $50,000 for single retirees.)
At the same time, the SSA cut over 8,000 workers (13% to 14% of its staff) in the last year, including more than 3,800 customer service representatives, leaving the agency with “fewer employees than at any time since 1967,” according to the CBPP.
-
Could your generic medications get more expensive? Trump’s new tariff plan raises big questions
President Donald Trump is threatening to impose a 100% tariff on imported generic medications, which could raise costs for Americans unless drugmakers can move production to the U.S. by August 2028.
In his latest swipe at the pharmaceutical industry, Trump posted on Truth Social Tuesday that imported generic medicines would be subject to a 100% tariff beginning in August 2028 that would rise to 200% in August 2029. The president’s goal is to reshore production “to protect the people of the United States,” even though American consumers would likely end up paying much higher costs, as generic drugs make up more than 90% of prescriptions.
But experts caution that, as with some of Trump’s past tariff threats, there are more questions than answers right now about whether that timeline is feasible for reshoring production for drugmakers, how the tariff would be applied, and whether the president is willing to negotiate. What’s more, the Trump administration has yet to make a formal executive order or issue an official policy implementing tariffs on generic medications.
That’s why more details are needed, according to the Association for Accessible Medicines, which represents generic drugmakers. And there may be alternative legislative and regulatory solutions that could help address market deficiencies, AAM’s president and CEO John Murphy III said in a statement to Politico.
“We need to understand more the specifics of the policy, but the generics industry is committed to pursuing policies that support and stabilize both the industry and the access necessary to ensure patients have reliable options for affordable medicines,” Murphy said.
Still, the threat of tariffs has been enough to unnerve traders. Shares of various European and Asian companies that manufacture generic drugs fell as much as 4.3% on Wednesday, The Wall Street Journal reported.
WILL THIS MEAN HIGHER COSTS FOR CONSUMERS?
Whether idle or real, Trump’s latest tariff threat could upend the generic drug industry—and materially shock a current regime that’s ensured these medicines are more affordable in the U.S. than in many other countries, as Jeremy Leonard, managing director of global industry services at Oxford Economics, told Bloomberg. “The main effect is likely to be higher costs and supply disruption rather than a quick move of production to the U.S.”
While drugmakers enjoy years of exclusivity for brand-name drugs to account for the years of costly research and development that allow them to charge higher prices for the medicines, that’s not a luxury shared by generic drugmakers. Rather, after patents have expired and a proven market has been identified, makers of generic medicines enter the market and compete largely on price.
Thinner margins for generic drugs mean that tariffs of 100% or 200% would be very difficult for these drugmakers to absorb—and they’d likely have to pass some of those costs along to consumers or stop selling some products altogether, as CNBC reported. “A 100% to 200% tariff on a product with single-digit margins is a market-exit notice,” Salil Kallianpur, an independent pharmaceutical consultant, told the outlet.
If the supply of available generic drugs dries up for American consumers, that would increase costs by forcing them to rely on brand-name drugs instead. As a result, tariffs on generic drugs could end up having the opposite effect than Trump has said he wants to achieve, Nathan Gray, a senior research fellow at the Institute for International Trade at Adelaide University in Australia, told Bloomberg.
“If they want to reduce costs for consumers, this is not the way to do it,” Gray told the outlet.
WILL THE DRUGMAKERS MOVE PRODUCTION?
There’s also the question of whether Trump’s timeline for reshoring drug manufacturing is even realistic, as building new facilities may take longer than a couple of years.
Trump said in his post that such facilities are being built across the U.S. “at a level never seen before,” though he didn’t provide any specifics regarding that activity. In April, Trump announced a 100% tariff on certain brand-name and patented drugs to take effect on July 31 if drugmakers did not build factories in the U.S., though several drugmakers were able to avoid those tariffs by negotiating lower prices. Trump said on Tuesday that the policy “has been so successful” that it will remain in place.
But because of the economics of generic drugs, many companies may have very little incentive to move manufacturing to the U.S. The Indian Pharmaceutical Alliance, a trade group that represents 23 generic drugmakers, told The Wall Street Journal that tariffs alone wouldn’t lead to sustainable onshoring to the U.S. of generic-medicine production.When the Trump administration started floating the idea of pharmaceutical tariffs last year, the CEO of Sandoz, which makes more than 400 generic drugs, told The Wall Street Journal that such threats wouldn’t induce change—unless the government was willing to offer some help.
“Where’s the incentive?” Richard Saynor told the outlet. “You sell a packet of antibiotics more cheaply than a packet of M&M’s. That’s offensive, and we lose money doing that.”
-
Is your AI as good as it says it is?
AI has become very good at passing the tests we set for it.
Stanford University’s 2026 AI Index Report captures the problem neatly. While AI models can master abstract logic, they often struggles with basic spatial reasoning tasks. For instance, a leading AI model could win gold at the International Mathematical Olympiad, yet it could correctly read an analogue clock only half the time.That is the paradox of AI. Exceptional in one domain. Unreliable in another.
That unevenness matters because AI is rarely marketed as conditional, only as capable.
We have seen the same pattern play out on some of the biggest stages in tech. For instance, Tesla’s Optimus robots were presented as a glimpse of autonomous robotics, but after their appearance at a 2024 event it was reported they relied on human intervention for some capabilities. Meta’s AI glasses failed twice during live demos, with the company later pointing to technical issues.
The details differ, but the pattern is consistent: AI can look capable in controlled conditions and then behave very differently when it meets the messiness of the real world.
The same gap is showing up across industries. MIT’s Project NANDA Gen AI study found that 95% of organizations are getting zero return, with most systems stuck without measurable P&L impact.
The question worth asking is simpler than it sounds: is your AI being measured on your reality, or someone else’s?
THE AGE OF BENCHMAXXING
Every competitive industry eventually learns to optimize for its scorecard. AI is no different. Now there is even a name for it: benchmaxxing.
Benchmarks exist for good reasons. They create a common language. They make comparison possible. The problem emerges when the benchmark becomes the target rather than the measurement. Once that happens, optimizing for a test and genuinely improving capability become two different activities.
Benchmarks themselves aren’t as solid as they seem. A 2025 study found that giving developers even limited access to test data could boost leaderboard scores by up to 112%. Meanwhile, a February 2026 paper found nearly half of widely used benchmarks have hit saturation, meaning top models score so similarly that the tests can no longer distinguish between them.
In my corner of the industry, I hear the same claims multiple times a day. World’s best. Fastest. Most accurate. Vendors are finding ever more creative ways to outdo each other on whatever metric is currently in fashion. What’s less visible is what those comparisons leave out: the models that didn’t make the cut, the test conditions that weren’t disclosed, the user populations that were never included in the first place.
A benchmark can tell you how a model performs on a test. What it cannot tell you is how that model performs under your conditions, with your users, on the problems that actually matter to your business.
THE SHOWROOM AND THE ROAD
Demos are designed to showcase strengths. That means, by definition, removing the conditions that create problems in production: controlled environments, predictable inputs, carefully chosen use cases, users who behave exactly as anticipated. Nobody demos the edge cases.
The gap this creates is well documented and almost universally underestimated. Recent analysis confirms it:most teams discover the hard way, after a prototype that dazzled stakeholders starts silently degrading in production. In my space—voice AI—the metrics chosen for demos are part of the same problem. Vendors lead with the numbers they’re confident winning on. The measures that would reveal weaknesses, how a system performs under pressure, with difficult inputs, at scale, tend not to make it onto the slide.
Demos are not dishonest. But they are incomplete by design, and buyers rarely have enough information to know where the demonstration ends and the actual product begins.
WHO WAS THIS BUILT FOR?
Benchmark saturation isn’t the only blind spot; representation is the other.
Every evaluation framework makes choices about who it includes. Those choices determine whose experience is measured, who it’s optimized for, and who is quietly treated as an edge case.
This is a huge sticking point in voice AI. A system might perform well in a clean test set and still struggle with the way people actually speak: regional accents, dialects, code-switching, overlapping speech, background noise, interruptions, mumbling, laughter, emotion, technical language, older speakers, second-language speakers.
The risk is that a buyer sees a single accuracy number and assumes it represents everyone they serve. It rarely does. A model trained and tested on narrow conditions will look strong for the people most represented in that data. It may perform very differently for everyone else.
That gap does not always show up as an obvious failure. It shows up as more corrections, more friction, more abandonment, more human review, and worse outcomes for the users least visible in the evaluation.WHAT TO DO ABOUT IT
Benchmarks are useful signals. Treat them as a starting point, not a conclusion.
Before any procurement decision, ask vendors to demonstrate performance on your specific use cases, with your actual user population, under conditions that resemble production rather than a demo script. If they cannot provide it, that gap in the evidence is itself the answer.
Test against edge cases deliberately. The users most likely to be underserved by a system are rarely the ones centered in the demo. Include them in evaluation from the start.
Then keep measuring after deployment. AI performance is not a fixed point. It degrades, drifts, and surprises.
The organizations extracting real value from AI are not the ones that ran the best procurement process. They are the ones that treated go-live as the beginning of evaluation, not the end of it.
The question is no longer whether AI can pass the test. It is whether the test resembles reality.
Katy Wigdahl is CEO of Speechmatics.
-
Chick-fil-A says hackers accessed rewards accounts. Here’s what may have been exposed
If you’ve eaten at Chick-fil-A lately, your spicy chicken sandwich might have come with something extra—and it isn’t waffle fries.
The country’s most popular chicken sandwich chain just got hit with a data breach targeting its most loyal customers. Chick-fil-A disclosed the incident this week, noting that it identified “suspicious login activity” on some of its Chick-fil-A One rewards program accounts last month.
After conducting an investigation, the fast-food chain determined that hackers mounted an automated attack against its website and app for three days in mid-June, using a list of email logins and passwords obtained through a third party. The company is letting customers know that the unauthorized users may have gained access to data stored in their Chick-fil-A One rewards program accounts, including names, email addresses, phone numbers, birth dates, addresses, and the last four digits of credit card numbers.
“Chick-fil-A takes the protection of personal information seriously,” the company said in a statement published to its website. “As soon as Chick-fil-A discovered the incident, we immediately took action to protect customers’ accounts, which included forcing log-outs of affected accounts and removing any stored payment methods. We also restored impacted customers’ Chick-fil-A One account balances.”
One login to rule them all
The technique used in the Chick-fil-A breach, known as credential stuffing, happens when hackers leverage a large list of stolen username and password combinations to see what other accounts they can gain access to. Because people reuse passwords, one set of credentials often unlocks unrelated accounts—a good reminder not to share your chicken sandwich loyalty program password with the account you use to monitor your 401(k).
Based on the notification letters from the company, the breach appears to have affected Chick-fil-A customers in Iowa, Maryland, Massachusetts, New Mexico, New York, North Carolina, Oregon, Rhode Island, Vermont, and Washington, D.C. The company is encouraging customers who belong to its loyalty program to update their passwords to something unique, a step that can prevent hackers from using stolen databases to easily crack open online accounts.
Chick-fil-A encourages customers to sign up for its loyalty program to order ahead for pickup and to reap rewards like free sides and desserts. Through its app, loyalty program members collect points with their purchases that they can spend toward free items in the future. Loyalty apps have exploded in the fast-food space in recent years as a way for brands to boost retention and capture more information about customers and their behavior. Fast-food chains can also target discounts and special offers specifically to customers willing to sign up and hand over some data—a juicy offer with inflation cutting into even the cheapest tier of restaurant experience.
Much like the rest of the data we give up online, loyalty programs come with trade-offs. Chick-fil-A is reassuring customers that its breach is now handled, but no data is safe in the digital world—not even your stash of chicken sando points.
“As an additional way to say thank you for being a loyal Chick-fil-A customer, we have added rewards to your account,” the company said in its letter notifying customers. “Chick-fil-A continues to enhance its security, monitoring, and fraud controls as appropriate to minimize the risk of any similar incident in the future.”