Sam Altman appeared virtually at the Commonwealth Bank of Australia's Sydney conference on May 26, 2026 , and publicly reversed one of his most-cited predictions: that AI would rapidly hollow out entry-level white-collar work. "I'm delighted to be wrong," he told CBA CEO Matt Comyn. The correction lands against a live backdrop of Cisco, Meta, and Standard Chartered announcing AI-linked layoffs this year — and a Gartner finding that 80% of executives have already cut headcount to fund AI investment .
What Altman Said — and What He's Walked Back
After ChatGPT's launch in November 2022 , Altman warned that customer support roles would disappear quickly and predicted AI could compress a job-turnover cycle that historically moves 50% of roles over roughly 75 years into a drastically shorter window. His updated assessment, stated at the CBA event, walks that back across the board.
"I'm delighted to be wrong about this. I thought there would have been more impact on entry-level white-collar jobs being eliminated by now than has actually happened. I now think I understand more about why it hasn't, and I'm obviously grateful — but that is an area where my intuitions were just off." — Sam Altman, CEO at OpenAI, CBA Sydney conference, May 26, 2026
His direct framing on the forecast: "I don't think we're going to have the kind of jobs apocalypse that some of the companies in our space advocate or talk about." OpenAI's record, he said, was "roughly right on the technology itself" but "pretty wrong on social and economic consequences" . He stopped short of a full reversal: "At the time I was like, I see this is a real risk we should probably talk about. And it still may" . The update is on timeline and magnitude, not structural certainty.
The Slack Experiment: What Updated His Model
The position shift came partly from a direct test. Altman configured an AI agent to respond to his Slack messages and email, announcing to recipients "this is Sam's AI." The agent was technically capable. He reverted to handling messages himself anyway.
"We really do care about our interactions with people. This thing… is not something that I can imagine myself outsourcing to an AI anytime soon." — Sam Altman, CEO at OpenAI, May 26, 2026
The theoretical update: jobs contain a "human part" — relational and social dimensions — that creates a meaningful adoption ceiling even where an AI agent is technically adequate. For developers building on LLMs, this is a design constraint, not just an economic observation. Agents that silently replace human touchpoints may encounter organizational friction that capability benchmarks don't predict. Human-in-the-loop patterns and explicit handoff design are practical responses to this ceiling.
AI Washing: Layoffs Blamed on AI That Would Have Happened Anyway
At India's AI Impact Summit in February 2026 , Altman introduced the term "AI washing": companies attributing headcount reductions to AI when the actual driver is ordinary cost pressure. The concept complicates any attempt to read layoff announcements as a clean displacement signal.
"There's some AI washing where people are blaming AI for layoffs that they would otherwise do, and then there's some real displacement by AI of different kinds of jobs." — Sam Altman, CEO at OpenAI, CNBC-TV18, February 2026
He didn't frame "AI washing" as exculpatory. His broader forecast remains: the "real impact of AI doing jobs in the next few years will begin to be palpable," with net new job creation following prior tech-revolution patterns . No measurable timeline or scale was offered for either prediction. For developers reading macro signals: an uptick in AI-attributed layoffs isn't a clean displacement input. Organizations have structural incentives to invoke AI as a modernization narrative regardless of actual cause.
What's Actually Happening in the Labor Market: 2026 Data
Altman's revised optimism runs against a set of concrete 2026 numbers. Several large employers have announced AI-linked workforce reductions this year, and a Gartner survey captures executive behavior at scale.
| Company | 2026 Action | Stated AI Link |
|---|---|---|
| Cisco | ~4,000 layoffs | AI investment reallocation |
| Meta | ~8,000 roles cut | AI investment reallocation |
| Standard Chartered | ~8,000 back-office roles flagged | AI replacement of back-office functions |
| HSBC, Amazon, CBA | Workforce reshaping acknowledged | AI-driven efficiency initiatives |
| Gartner executive survey | 80% cut staff to fund AI | Direct AI funding trade-off |
The Gartner figure is the sharpest empirical point behind the "AI washing" framing. It confirms that AI is already altering headcount decisions — while simultaneously muddying attribution, since "cut staff to fund AI" is mechanically different from "replaced by AI." The Yale Budget Lab, using BLS Current Population Survey data through March 2026 , found no statistically significant differences in occupational change or unemployment duration for workers in high AI-exposure roles since ChatGPT's launch. Consistent with Altman's observation — but a lag indicator, not a forward signal.
Where Altman and Anthropic Diverge
The revised view is not the consensus position inside AI labs. In the same week, Anthropic co-founder Christoph Olah stated there remains "a real possibility" that AI could displace human labor "at very large scale" — a direct counterpoint to Altman's softer framing.
The divergence has a structural dimension. OpenAI's commercial position creates incentives to reduce organizational friction around AI adoption; a softened displacement narrative helps enterprise procurement cycles. Anthropic's safety-first public stance produces a different incentive: acknowledging displacement risk aligns with its stated mission. Neither position is neutral on the question of how displacement risk should be communicated.
Neither camp offers falsifiable forecasts. There are no published thresholds — no "if X% of white-collar roles are eliminated by Y date, the apocalypse framing was correct." Both positions are effectively unfalsifiable on their own terms, which is relevant context when either view is used to inform tooling investment or workforce planning decisions at the technical founder or team level.
What to Watch Next
Altman's hedge — "it still may" — means the all-clear is conditional. His update is to timing and magnitude. Several indicators will resolve the uncertainty more clearly than any executive statement will.
- BLS Occupational Employment and Wage Statistics (quarterly): The most direct signal on occupational churn in high AI-exposure roles. A sustained deviation from historical baselines would shift the empirical picture.
- JOLTS quits rate in white-collar sectors: A rising voluntary-quits rate among administrative and professional workers — particularly in roles with high AI-task overlap — would be an early structural displacement indicator.
- EU AI Act disclosure obligations: If AI-linked layoffs accelerate, expect regulatory responses in the EU and potentially the US that could constrain enterprise AI deployment timelines and affect procurement decisions for LLM-based products.
- Agent adoption friction data: Altman's "human part" insight is testable. Teams building on LLMs should track whether human-in-the-loop flows outperform fully automated flows on adoption metrics — particularly in workflows with strong relational components. The outcome will either validate or falsify the "human part" as a design constraint.
Frequently Asked Questions
Why did Sam Altman change his position on AI job displacement?
Two factors combined. First, observed evidence: entry-level white-collar roles haven't disappeared at the rate he predicted after ChatGPT's November 2022 launch . Second, a personal experiment: he routed his Slack and email through an AI agent and found the relational dimension of communication couldn't be handed off, even when the AI was technically functional. He now believes many jobs contain an irreducible "human part" that limits displacement independently of capability levels .
What is 'AI washing' in the context of layoffs?
Altman's term — introduced at India's AI Impact Summit in February 2026 — for companies attributing headcount reductions to AI when the actual driver is ordinary cost pressure. The Gartner finding that 80% of executives cut staff to fund AI illustrates the measurement difficulty: AI is clearly affecting headcount decisions, but "to fund AI" is mechanically different from "replaced by AI," and the two categories don't separate cleanly in public announcements.
Does Altman rule out future AI-driven job losses?
No. His explicit hedge at the CBA conference: "it still may" happen. The revision concerns timing and magnitude, not a structural reversal of displacement risk. He still expects the "real impact of AI doing jobs in the next few years will begin to be palpable" and forecasts net new job creation following prior tech-revolution patterns — without specifying measurable criteria for either prediction.
Are real companies already cutting jobs because of AI?
Yes. In 2026 alone, Cisco announced approximately 4,000 layoffs, Meta cut around 8,000 roles, and Standard Chartered flagged nearly 8,000 back-office positions for AI replacement . HSBC, Amazon, and CBA have also acknowledged AI-driven workforce reshaping. Altman would note that some portion reflects "AI washing," but the Gartner aggregate data — 80% of executives linking staff cuts to AI investment — points to real behavioral change at scale.
What does the 'human part' of jobs mean for developers building AI agents?
Capability benchmarks are an incomplete predictor of real-world adoption. Even technically proficient agents may face organizational friction in workflows with strong relational or social dimensions. Altman's Slack experiment is the concrete illustration: his AI was functional; he reverted anyway . For teams building on LLMs, this means human-in-the-loop patterns and explicit handoff design are practical requirements, not optional refinements — particularly in communication, support, and relationship-management workflows.
Where This Leaves the Forecast
Altman's public reversal is the most prominent shift in AI-leadership positioning on labor risk since the post-ChatGPT alarm cycle began. It's grounded in something real: predicted mass displacement of entry-level white-collar work hasn't materialized at scale through May 2026 . But the Gartner numbers, the Cisco and Meta headcount announcements, and Christoph Olah's direct counter-reading all argue the picture is more contested than a single revised forecast resolves.
The "human part" framework is the most practically useful output from Altman's updated view. It reframes adoption risk from a capability question to a design question. Teams building on LLMs have a concrete response available: design agents that augment relational workflows rather than replacing them outright. That hedge holds value independent of whether the macro displacement narrative ultimately proves right or wrong — and it's testable in production without waiting for the labor statistics to catch up.
Last updated: 2026-05-26. Based on Altman's remarks at the CBA Sydney conference and his February 2026 CNBC-TV18 interview; cross-referenced with Euronews reporting and Yale Budget Lab BLS analysis. Article reviewed against original source transcripts.
