Italy published Legislative Decree 179/2026 in its Official Gazette on Thursday, October 8, and one article in it matters to any board with staff there. From October 23, an employer that uses AI in decisions about hiring, changing or ending a job, including discipline and performance reviews, cannot let the system make the final call. That call belongs to a person with real, independent authority, and a worker's consent does not waive it.

Employers have to tell workers before the processing starts. On request, a worker gets a plain explanation of the decision from a human, covering what part AI played and the main factors it weighed. A dismissal that breaks the rule is void, and so is any other decision made the same way. One carve-out: searching for and screening candidates is not a final hiring decision, even when it ends someone's application. (Gazzetta Ufficiale; DirittoBancario summary)

The rest of the decree is the plumbing for the EU AI Act in Italy. The National Cybersecurity Agency (ACN) becomes the main market surveillance authority and the single point of contact with Brussels. It can open cases on serious incident reports, complaints, referrals from other bodies or on its own initiative, and it can inspect real-world AI tests remotely or on site without notice. The Agency for Digital Italy (AgID) oversees the bodies that certify high-risk systems.

Banks, investment firms and insurers answer to their usual supervisors: the Bank of Italy, CONSOB and IVASS take charge of high-risk AI used in lending, financial services and insurance. Fines track the AI Act: up to €35 million or 7% of global turnover for banned practices, up to €15 million or 3% for breaking provider and deployer duties, and €7.5 million or 1% for feeding regulators misleading information. For minor breaches, an authority can order a fix or publish the company's name with the violation instead. (Gazzetta Ufficiale)

The timing is the board point. Brussels pushed the AI Act's high-risk rules for areas like employment out to December 2, 2027. Italy's human-decision rule for workers starts in two weeks, so a company running AI in its Italian people processes cannot wait for the EU date. The decree also follows a September decree on civil liability for harm caused by AI.

Professionals have a clause of their own. Under Article 49, use of AI can change the fair fee that a lawyer, accountant or other professional is owed, with parameters tied to the risk class of the system, to be set within twelve months. The public administration minister will also set minimum AI skills for civil servants, and the decree creates a national AI sandbox run jointly by AgID and ACN. (AI Act Service Desk; Gazzetta Ufficiale)

In New Delhi on Thursday, IT Minister Ashwini Vaishnaw said India will publish a consultation paper on AI regulation within 30 days, covering safety, security, skills and what he called a human-first approach. He said AI cannot be regulated by a law alone, and that industry will carry much of the load. India has no single AI law today; most AI tools fall under the Information Technology Act of 2000.

In the US, USA Today Co. and the local papers it owns, among them The Tennessean, IndyStar and The Columbus Dispatch, sued OpenAI in a filing on Thursday. The publisher says OpenAI copied hundreds of thousands of its articles to train its models and is seeking more than $250 million in damages. It joins a long list of publishers already in court with OpenAI, The New York Times among them. (Business Standard; The Verge)

Yesterday's The Ten stays sent: the UK Information Commissioner's Office and the ten model developers that changed their data practices after two years of supervision. Thursday's The Guidelines stays sent: the Monetary Authority of Singapore's final AI risk-management Guidelines for financial institutions. Wednesday's The Airlock stays sent: the UK accepted all 44 National Commission recommendations on AI in healthcare. Tuesday's The Force stays sent: the Super Intelligence Force. Monday's The Vision stays sent: the Kyoto Vision for a Golden Age of Science. Sunday's The Clause stays sent: General Services Administration clause 552.239-7001. The Probe (October 2) and The Accord (October 1) stay sent too.

Last Saturday's The Subpoena also stays sent. Read today next to The Ten: London said autonomy is no excuse for poor compliance, and Rome now says a machine cannot sign off on a person's job. The older stack, from Florida's Hold to Agents That Probe, stays where it is. Today's file is the Decree. (The Ten; The Guidelines; The Airlock; The Force; The Vision; The Clause; The Probe; The Accord)

The Agencia Española de Protección de Datos (AEPD) notice on a breach carried out by an AI agent, the Korea Internet & Security Agency (KISA) red lines for autonomous AI, Altman's monitorability warning and the multi-lab pact talk, and Amodei's case for pacing the frontier all stay on the desk.

Madrid is still the one to reread. A European data regulator has already treated an agent-run attack as a personal data breach, and Rome now has a cybersecurity agency that can inspect AI tests without notice. (AEPD; SEDaily; TechTimes; Amodei)

If you employ people in Italy, put three blanks on one page before October 23. Where does any AI tool touch hiring, promotion, discipline, performance ratings or dismissal? Who is the named person who makes the final call in each case, and do they have real authority to overrule the system? If a worker asks why, can that person explain the decision in plain language, including what the AI contributed?

Then check the vendors. If the HR software cannot tell you which factors drove a score, nobody on your side can give the explanation the law will require from October 23. A void dismissal costs more than an audit. (Gazzetta Ufficiale; DirittoBancario summary)

The Move

Italy now requires a person to make the final call in every AI-assisted employment decision, from performance reviews to dismissals, starting October 23, and given workers the right to have a human explain it. Break the rule and the dismissal is void.

Ask the head of HR and employment counsel for a one-page map: every AI tool in your Italian people processes, the person who makes the final call, and how you would explain a decision if a worker asked tomorrow.

If the answer is "we only use it for screening," check that it really stops there. Screening is carved out. Ratings and firing are not.

Seven bullets. One move.