Technology & innovation
As AI becomes more deeply embedded in recruiting, branding executive Claire Bahn is raising a concern that's getting less attention than the efficiency gains: the more candidates optimize for AI-driven hiring, the more interchangeable they become.
Recruiting is now the top HR function where organizations are using AI, according to SHRM's 2025 research.
19% of organizations using AI in hiring said their tools had overlooked or screened out qualified applicants.
Bahn argues that AI-shaped hiring is flooding candidate pools with polished, algorithm-friendly applications that are increasingly hard to distinguish from one another.
The concern: AI may be rewarding optimization over judgment, making it easier to move the wrong candidates forward with more confidence than they've earned.
Read more via HR Executive
Unrepresented plaintiffs are filing more employment lawsuits than ever, and generative AI is making those cases more expensive to defend even when employers almost always win.
Pro se filings in federal employment cases more than doubled between 2021 and 2025, rising from 2,052 to 4,388 cases, and their share of all federal employment litigation grew from 9.7% to 16.5%, according to LexisNexis Lex Machina's 2026 Employment Litigation Report.
Federal employment lawsuits hit a record 26,635 in 2025, the highest total in at least a decade.
Employers still prevail by a margin of more than 40 to 1 in pro se cases, but AI is keeping weak cases alive longer and driving up defense costs, in part because AI tools are generating inflated settlement demands often in the hundreds of thousands of dollars.
Courts are beginning to push back, imposing sanctions ranging from $500 to $10,000 for filings containing fabricated citations, and in at least one case relieving an employer of any obligation to respond to future filings after a plaintiff filed more than 49 AI-assisted motions.
Read more via Baker Donelson
New research suggests the U.S. is building a lead in AI adoption that could translate into a productivity advantage similar to the one it gained during the technology boom of the 1990s.
43% of U.S. workers reported using AI for their jobs in early 2026, the highest share among seven countries surveyed by Brookings researchers. Adoption in European countries ranged from 36% in the U.K. down to 26% in Italy.
The gap is wider than those numbers suggest. U.S. workers don't just use AI more often, they use it more intensively. About 5% of all U.S. work hours in early 2026 involved AI, roughly double the share in the U.K. and more than triple the share in Germany, France and Italy.
Within Europe, northern countries lead. Norway tops the continent at 56%, followed by Denmark at 48% and Finland at 46%, according to Eurostat. Romania ranks last, with fewer than one in five people reporting recent AI use.
At the company level in Romania, nearly half of employers use AI only sporadically, and usually because individual employees decided to on their own, not because the company has a policy, according to a new eJobs survey of 196 Romanian companies.
The productivity stakes are real. Brookings researchers found that a 10 percentage point increase in AI adoption is associated with nearly 3 additional percentage points of cumulative productivity growth. The current U.S.-Europe gap implies the U.S. has already pulled roughly 3 percentage points ahead since 2022.
So far, neither U.S. nor European data show AI adoption associated with job losses at the industry level.
Read more via Brookings/St. Louis Fed, Visual Capitalist, Business Review
As traditional entry-level hiring slows, a growing gig economy is recruiting college students to train AI models, paying anywhere from $23 an hour for general data annotation to $200 for specialized expert work.
Platforms like Mercor and Handshake are specifically targeting undergraduates, who one student described as appealing because they "have a lot of time, generally, compared to working class professionals."
The work involves post-training AI models, the phase that shapes behavior and alignment with human objectives, a task that requires human judgment and can't be automated the way pre-training data can.
The gig economy angle is complicated by the Mercor data breach last month, which exposed personal information of contractors working on exactly these kinds of projects.
Read more via The Daily Californian
Chinese humanoid robot beats human half-marathon world record: A humanoid robot called Lightning, made by Chinese smartphone maker Honor, won a Beijing half-marathon race in 50 minutes and 26 seconds, beating the human world record of 57:20. Lightning ran autonomously (aside from one human assist after it slammed into a barricade near the finish line). Honor robots swept the top three spots. Last year's winner, Tien Kung Ultra, more than halved its time from 2024. China shipped more than 1,000 humanoid robots last year from at least three companies; no American maker delivered more than 500. (The Wall Street Journal)
Anthropic is investigating unauthorized access to its Mythos AI model: A small group of users in a private online forum gained access to Claude Mythos Preview through a third-party contractor environment on the same day Anthropic announced its limited release, according to Bloomberg. The group has been using the model since, though not for cybersecurity purposes. Users gained access using a combination of contractor credentials and internet sleuthing tools, partly aided by information exposed in the recent Mercor data breach. Anthropic said it has no evidence the access extended beyond the third-party vendor environment or affected its own systems. (BBC, Bloomberg)
Meta is laying off 8,000 people, and it's not hiding the reason: The company plans to eliminate 10% of its workforce on May 20 as it offsets the cost of up to $135 billion in AI infrastructure spending this year. Employees also learned this week that Meta is rolling out a tool that tracks keystrokes, mouse movements, and clicks to train its AI models. There is no opt-out. Worker sentiment on anonymous employee platform Blind has turned sharply negative, with more than 80% of posts about Meta this year classified as negative, up from about 20% in 2024. (CNBC, The Wall Street Journal, The New York Times)
