On Friday evening Beijing time, China's State Council published a guideline on adapting to AI without wrecking employment. It sets out 18 tasks in six areas. The sharpest is aimed at employers: companies rolling AI out at scale are encouraged to set up an "unemployment buffer period," telling affected workers in advance and helping them retrain or move to new jobs. Beijing also wants AI used first where labor is short or work is dangerous, and a careful hand on the scope, strength and pace of AI in labor-intensive industries.

There is a gate as well as a nudge. AI projects in market-access areas that could cause a big shock to jobs will face an employment risk assessment during review, and AI-driven job losses are being added to China's system for managing large-scale unemployment risk. The guideline also targets AI in hiring and HR, naming algorithmic discrimination, fake recruiting and invasions of privacy and rest time. The world's second-largest economy now treats AI job losses as a risk to manage, with paperwork attached. (State Council General Office; English summary)

In New York the same day, the decision looked different. American Express said it will cut 5% of its global workforce and take a $336 million restructuring charge. Its finance chief, Christophe Le Caillec, told CNBC the point was to make the company "leaner, more agile, and also to prepare for AI," by removing management layers before automating the work underneath them.

Chief executive Stephen Squeri spent ten minutes of the earnings call on AI and drew one firm line: outside AI agents will reach Amex systems only through approved channels, because the company is "not going to let assistants run rogue through our system." The cuts came in a quarter when profit beat forecasts. No buffer period was announced. (CNBC)

For a look at where those jobs go, Anthropic has published a study of its own workforce. The junior share of its technical departments slipped from 21% to 14% since early 2025, and the junior share of technical hires fell faster, from 27% to 9%. By August, agents directed by individual employees were logging six times as many hours as the people directing them.

The work changed shape. Agents spend about 80% of interactive sessions waiting on a person, and the share of sessions spent managing agents, writing specs for them and building tools to monitor them, rose from 10% to 22%. Hiring managers now weigh judgment and autonomy above narrow expertise. If the entry rung is thinning at an AI lab, your own talent pipeline needs a look. (Anthropic)

Seoul is spending its way through the same transition. On Friday Kwon Dae-young, South Korea's vice minister of economy and finance, said the government's budget proposal more than triples spending on physical AI, from 900 billion won this year to 3.1 trillion won (about $2.3 billion) next year, to put AI into factories and robots.

He paired the money with a promise. A free AI service for everyone is due by the end of this year, and the AI education budget rises to 2.1 trillion won. "AI must not become an opportunity for some and a wall for others," Kwon said. Two Asian governments, one Friday, both putting jobs inside AI policy. (The Herald Business)

Already filed, if you want the run-up: The Lab (Britain's new AI biosecurity lab, funded partly in OpenAI and Anthropic credits), The Clerk (a US appeals court taking a case away from a judge over an AI-drafted order) and The Tab (Europe's Parliament fighting over who pays when AI does harm).

Also on the shelf: The Footer (Brazil's top court fining lawyers for hiding orders to its AI), The Loop (Japan's rulebook for 180,000 civil servants using AI agents) and The Stake (Anthropic putting 5% of its shares into a public benefit fund). Read today beside The Loop. Japan wrote rules for public servants who work with AI. China is now writing them for the workers AI might replace.

The idea of warning workers first is not new. In 1988 Senator Howard Metzenbaum of Ohio introduced the Worker Adjustment and Retraining Notification Act. It passed the Senate 72 to 23 and the House 286 to 136, and President Reagan let it become law without his signature.

The WARN Act requires most employers with 100 or more workers to give 60 days' notice before plant closings and mass layoffs, so people have time to find new work or retrain. It was written for shuttered factories. China's buffer period applies the same logic to software. If AI cuts come in waves, the old notice rules may be the first test of how your company handles them. (WARN Act)

Put these to management. Where do we expect AI to remove roles in the next two years, and how many? Who decides the pace, and would the affected people hear it from us before they read it in an earnings call?

Then the pipeline. If junior roles shrink, where will our next managers come from? Do our layoff plans meet WARN and its foreign equivalents, and are our hiring tools checked for the kind of algorithmic bias Beijing just named?

The Move

China just wrote down something most boards have never discussed: whether workers hear about AI cuts before they happen. Amex took the other path the same day and announced its cuts on an earnings call. Both are choices. Staff, regulators and investors will notice which one you made.

Ask the chief people officer for one page before the next board meeting: which roles AI is likely to remove, on what timeline, what notice and retraining we will offer, and what that means for the junior pipeline.

Decide the buffer before you need it.

Seven bullets. One move.