Career Intel with Dan 📊 | The AI Builders Aren't Immune: This Week's Layoffs Hit AI's Own Research Teams

DEFINING PATTERN:

AI was the leading stated reason for US job cuts for a fourth consecutive month in June, per Challenger, Gray & Christmas, with tech-sector layoffs up 83% year-over-year to nearly 139,000 in the first half of 2026, and even Amazon's own AI-research unit wasn't spared this week.

But the response to that pressure is fracturing rather than converging: Colorado just loosened its AI-hiring law even as Connecticut tightened disclosure and the EU bought employers 16 more months to comply with its high-risk rules, while Congress used a field hearing in Georgia to argue for regional flexibility over a single federal mandate.

Underneath the layoff headlines, AI-skill demand is genuinely spreading into accounting, banking, and staffing, not just tech, suggesting real reskilling opportunity for workforce systems that move fast enough to catch it.

🔊 1. Amazon Cuts Jobs in Its Own AGI Unit While Pouring Billions Into AI Infrastructure

On July 22, Amazon confirmed layoffs within its Artificial General Intelligence organization, the unit building its foundation models, even as the company continues to pour money into AI infrastructure (reported at roughly $200 billion). Teams under VPs Adeeb Shanaa (AGI data services) and Vishal Sharma (AGI information) were affected; Amazon did not disclose a total headcount figure, though employees posting on internal forums described cuts of around 10% in some groups, including model customization and post-training roles. Affected US employees receive 90 days of pay and benefits, outplacement support, transitional healthcare, and severance eligibility. This adds to Amazon's roughly 16,000 corporate job cuts announced earlier this year, bringing total reductions to about 30,000 since October 2025. Source: HR Katha, Tech Times, The Register, BusinessToday.

🔹 AGI data services and AGI information teams, led by VPs Adeeb Shanaa and Vishal Sharma, saw the reductions, concentrated in model customization and post-training roles

🔹 Affected US employees receive 90 days of pay and benefits, outplacement support, transitional healthcare, and severance eligibility

🔹 This adds to roughly 30,000 Amazon job cuts since October 2025, even as the company commits an estimated $200 billion to AI infrastructure

💡 Amazon is one of the labs building frontier AI, and it is still cutting the people who build it, evidence that no corner of the AI economy, including AI research itself, is insulated from restructuring toward deployment over research.

Impact: Immediate, ongoing.

🔊 2. Update: Judge Orders Meta to Explain Visa Holder Selections in Its AI Layoff Lawsuit

The Meta AI-layoff lawsuit covered here last week escalated on July 17, when US District Judge William Orrick declined to block Meta's roughly 8,000-person layoff round but found that four plaintiffs on Meta-sponsored work visas face immigration harm that is "more than mere speculation." Orrick ordered Meta to explain, by July 23, how and why those four employees were selected for termination. The court also found the 26 plaintiffs, current and former Meta employees who allege Meta used AI-generated ratings, productivity data, and internal activity signals to select layoffs, raised "serious questions going to the merits" of their claims. A hearing on a preliminary injunction is set for August 24. Meta maintains workforce decisions were made by people, not AI. Source: IBTimes UK, Courthouse News Service, Lumen Law Center.

🔹 Four employees on Meta-sponsored work visas are the immediate focus, their immigration status now tied to a live court order

🔹 Meta must explain its selection process for those four employees by July 23; a preliminary injunction hearing follows on August 24

🔹 The 26 plaintiffs allege Meta combined AI-generated performance ratings with productivity and activity data to select layoffs, a claim the court called serious enough to warrant scrutiny even while declining to halt the cuts

💡 This is shaping up as one of the first real court tests of whether AI-assisted layoff selection can be legally challenged, worth tracking as a bellwether for every employer using algorithmic tools in reduction-in-force decisions.

Impact: Immediate, active litigation; the legal precedent is longer-term.

🔊 3. Nearly Half of Hiring Managers Would Rather Invest in AI Than Hire a 2026 Grad

Two surveys released this week sharpen the picture on AI and entry-level hiring, and they do not fully agree with each other. A ResumeTemplates.com survey of 1,000 US hiring managers found 48% would rather invest in AI tools than hire and train a 2026 college graduate, rising to 65% in tech and 56% in finance; 55% of companies have already shifted entry-level hiring budget toward AI, and 45% have restructured so one senior employee using AI covers work that used to take several entry-level hires. Separately, Indeed Flex surveyed roughly 2,000 US and UK workers in June and found 62% believe companies are cutting entry-level hiring because of AI, but only 31% say they have actually seen that reduction happen in their own field. Source: ResumeTemplates.com via PR Newswire, Indeed Flex.

🔹 Government hiring managers are the outlier, just 20% would pick AI over a grad, versus 65% in tech and 56% in finance

🔹 23% of surveyed hiring managers plan to cut 2026 grad hiring outright, taking fewer or none

🔹 Only 31% of workers report seeing entry-level cuts firsthand versus the 62% who believe it is happening broadly, a real gap between perception and local experience

💡 The employer-side data (ResumeTemplates) and the worker-perception data (Indeed Flex) point to the same underlying shift, but the size of the gap between belief and lived experience suggests some of the entry-level displacement narrative is running ahead of what is actually visible on the ground. Note for regular readers: this is the third recent edition to touch the entry-level squeeze theme, following PwC's seniorized-jobs data on 07/04 and 07/11, the underlying pattern keeps recurring even as the specific sources change.

Impact: Immediate for hiring managers now planning fall 2026 grad recruiting; emerging as a research narrative.

🔊 4. Colorado Quietly Weakens Its AI Hiring Law While Connecticut and New York Tighten

State AI-employment law took a less uniform turn this week than some coverage suggests. Colorado Governor Jared Polis signed SB 26-189 on May 14, and it drew fresh attention this month: the law actually narrows Colorado's original AI-employment framework, repealing the mandatory risk-management program, annual impact assessments, and self-reporting duties that were coming under the original 2024 law, in favor of a lighter notice-and-disclosure regime for automated decision-making technology, effective January 1, 2027. That is a real reduction of employer obligations, not a tightening. Connecticut moved the opposite direction: its CART Act, signed May 29, requires employers to flag AI-related layoffs on WARN notices starting October 1, 2026, and to give pre-decision notice of automated employment tools starting October 1, 2027. New York City's Local Law 144 remains the only US law that mandates a bias audit by name, but a December 2025 state comptroller audit found significant enforcement gaps. Source: Consumer Finance Monitor, Buchalter, Littler, Nelson Mullins, Warden AI.

🔹 Colorado's rewrite removes the NIST AI RMF-aligned risk management program and 90-day impact assessments that would have applied to employment ADMT

🔹 Connecticut's CART Act ties AI use directly to WARN Act layoff notices starting October 1, 2026, a new disclosure mechanism other states have not adopted

🔹 NYC's Local Law 144 is still the only law that names a mandatory bias audit, but the state comptroller's own audit found the city is not enforcing it consistently

💡 The "states are marching toward stricter AI hiring rules" framing oversimplifies what is happening. Colorado just loosened its law, Connecticut just tightened disclosure, and even NYC's flagship audit mandate has enforcement gaps, so employers face a genuine patchwork, not a single national trend line. Correction note: earlier research drafts for this edition described Colorado's law as adding mandatory algorithmic audits; the opposite occurred, and that claim has been corrected here.

Impact: Immediate compliance work for multi-state employers; Colorado's lighter regime takes effect January 1, 2027.

🔊 5. Congressional Field Hearing in Augusta, Georgia Signals a Preference for Regional AI Workforce Policy

On July 24, the House Committee on Education and the Workforce held a field hearing in Augusta, Georgia, "Building an AI-Ready America: How AI Is Creating Opportunities Across America's Workforce," hosted by Rep. Rick Allen (R-GA) at Augusta University's Georgia Cyber Innovation and Training Center. Witnesses testified on AI's effects on manufacturing, healthcare, and agriculture, and the committee's questioning leaned toward avoiding one-size-fits-all federal mandates in favor of funding regional reskilling programs. Source: Committee on Education & the Workforce, C-SPAN.

🔹 Held at Augusta University's Georgia Cyber Innovation and Training Center, signaling a regional cybersecurity and tech-training framing for the testimony

🔹 Testimony spanned manufacturing, healthcare, and agriculture, sectors outside the usual tech-layoff headlines

🔹 The committee's stated preference for regional flexibility over blanket mandates previews how Department of Labor reskilling funding priorities may get shaped

💡 Congress is signaling it wants workforce boards and regional training providers, not a single federal AI-jobs policy, to carry the reskilling load, directly relevant for organizations like RochesterWorks positioning for future federal funding asks.

Impact: Emerging; shapes future DOL and workforce-development funding priorities rather than immediate rules.

🔊 6. AI Skill Demand Spreads Beyond Tech Into Finance, Accounting, and Staffing

The Bipartisan Policy Center's AI and Workforce Navigator, using Lightcast job-posting data, reported this week that AI skill demand is no longer concentrated in tech. As of May 2026, the number of US job postings listing AI skills had more than doubled year-over-year, with professional-services fields, including accounting, banking, and staffing, now showing growth on par with or ahead of tech roles. About three-quarters of all AI-skill demand remains concentrated in three occupational groups: computing and mathematics, business and finance, and management. Source: Bipartisan Policy Center, AI and Workforce Navigator (Lightcast data).

🔹 Accounting, banking, and staffing firms are now posting AI-skill roles at a pace that rivals or exceeds tech postings

🔹 Three occupational groups, computing and math, business and finance, and management, still account for roughly 75% of all AI-skill demand, meaning the diffusion is real but not yet broad-based

🔹 The Navigator explicitly flags that federal and state workforce systems were not built to track this kind of fast, cross-sector skill shift

💡 For workforce boards, this is a concrete signal to widen AI-skills curriculum beyond IT and into professional-services tracks like accounting and banking now, before demand outpaces training pipelines the way it already has in tech.

Impact: Emerging; direct input for training-provider curriculum planning.

🔊 7. AI Identities Become the Fastest-Growing Enterprise Attack Surface, Straining Governance

Sophos's AI Security 2026 Report, released this week, found that AI identities, autonomous agents, assistants, and the OAuth tokens and API keys they use, are now the fastest-growing new attack surface inside enterprises, as companies grant agents privileged access to finance, code, and customer-support systems to replace manual labor. Sophos documented a campaign it tracks as STAC6994, in which attackers ran roughly a dozen AI agents inside a compromised network to write and test attacks against endpoint security tools. Governance is visibly trying to catch up: this week also brought Microsoft 365 Copilot policy-based controls for managing large numbers of workplace agents, a compliance-focused prebuilt agent catalog from Squirro for finance, HR, legal, sales, and IT, and an operational risk-assessment playbook from Anthropic for security teams evaluating agentic AI. Source: Sophos AI Security 2026 Report, AI Agent Store weekly roundup.

🔹 Attackers are compromising OAuth tokens, AI service credentials, and exposed AI infrastructure, not just traditional endpoints, to gain network access

🔹 One documented attacker campaign, STAC6994, used about 12 AI agents inside a single compromised network to develop and test attacks

🔹 Microsoft, Squirro, and Anthropic all shipped agent-governance tooling in the same week, evidence vendors see this as an urgent, current gap rather than a future risk

💡 As companies replace manual labor with autonomous agents that hold real system access, the labor-market implication is not just fewer entry-level roles, it is rising demand for AI governance, auditing, and identity-security specialists, a growth lane worth flagging to IT and security career-track clients.

Impact: Immediate operational risk; growing demand signal for governance and security roles.

🔊 8. EU Delays AI Act's Workplace Rules to December 2027, But Employers Are Told Not to Relax

The European Parliament and Council finalized a delay to the EU AI Act's high-risk compliance deadlines this summer, an agreement known as the AI Omnibus that the Parliament approved June 16 and the Council cleared June 29, with publication expected in the EU Official Journal this month. For AI systems used in employment decisions, hiring, promotion, termination, task allocation, and worker monitoring, the compliance deadline moves from August 2, 2026 to December 2, 2027. The underlying high-risk obligations do not change, only the timeline. Source: Ogletree, Morgan Lewis, Fisher Phillips.

🔹 Hiring, promotion, termination, task allocation, and worker-monitoring tools are the specific employment use cases still classified high-risk

🔹 The compliance deadline moved 16 months, from August 2, 2026 to December 2, 2027, not an indefinite delay

🔹 Law firms advising employers are uniformly framing this as extra preparation time, not a reprieve from the underlying obligations

💡 US employers with EU operations now have more runway to audit hiring and monitoring AI tools against high-risk requirements, a good moment to start that work rather than wait, since the requirements themselves have not softened, only the clock.

Impact: Emerging; long-term compliance significance, immediate opportunity to prepare early.

BOTTOM LINE THIS WEEK:

The throughline this week is that AI-driven displacement and AI-driven opportunity are running on the same track, sometimes inside the same company. Amazon is cutting AGI researchers while still committing roughly $200 billion to AI infrastructure, Meta is defending algorithmic layoff decisions in court while facing a hearing next month, and hiring managers say they'd rather invest in AI than train new grads, even as workers' fears of entry-level cuts run well ahead of what most say they've actually witnessed.

Policy isn't offering a clean answer either, with states and the EU pulling in different directions on the same underlying question of how much scrutiny AI hiring tools deserve. For job seekers and workforce professionals, the practical read stays consistent with recent weeks: treat any single AI-layoff or AI-hiring headline as a data point to investigate, not a verdict, and prioritize building the judgment, coordination, and governance-adjacent skills that are visibly in growing demand even as headline hiring gets more selective.

Stay curious, stay current

Dan Lopez | danscareercorner.com