HBC-M018 · Italy — engineering conduct; Terni code example
Who keeps control when an engineer uses AI?
Compare AI assistance, professional judgment, client information and responsibility.
Published 15 November 2026
On this page
What the source says
Source summary
Italy’s Law 132/2025 Article 13 confines AI in intellectual professions to instrumental and support activities, with intellectual work predominant, and requires clear client information. This profile selects engineering: Terni’s local engineers’ body publishes a code ratified on 16 September 2026. The code’s Article 13 preserves independent judgment, requires advance information on AI characteristics and purposes, and retains responsibility, including for staff AI work. Missing information or improper use can attract discipline. This does not guarantee jobs, income or services without AI. Other territorial versions, later changes and individual coverage need checking.
Research date: 2026-11-15. See the sources, versions and relevant passages.
This profile examines specific roles and activities. It does not establish that an entire profession is reserved for people.
HBC analysis
Why preserve a human role here?
- Possible justification
- People commissioning consequential work may need a professional who can explain and defend the judgment, rather than merely accept a tool’s result. That reason depends on useful competence and actual authority; preserving a title or collecting a signature is insufficient.
- Strongest objection considered
- Requiring predominant professional input may protect incumbents, raise prices or obstruct effective assistance without improving the work. An undefined boundary can also leave professionals liable while clients or employers deny them time and information to challenge outputs.
- Who might benefit or be excluded
- Distinguish the individual professional, employed and junior engineers, clients, people affected by engineering work and those priced out of advice. Retaining professional control need not distribute productivity gains to employees or the public.
- Resources and costs to investigate
- Meaningful judgment needs time, competence, inspectable inputs and authority to reject unreliable work. Compare those costs with the risks and benefits of assistance; no budget, staffing level or net benefit has been estimated.
- An alternative to compare
- Compare a broad predominant-input standard with specified decisions that a professional must personally justify, supported by usable records and outcome monitoring. Test whether clients receive better explanations and work at an acceptable cost. No preferred model or trial result is claimed.
- What to check before using this elsewhere
- Identify the professional activity, person answerable, legal and code layers, territorial adoption, information owed and usable discipline or remedy route. Establish who can stop or revise an output, and who pays for that time, before proposing transfer.
- Evidence that could change the assessment
- Evidence that meaningful independent judgment improves work and client understanding at acceptable cost would strengthen the case. Routine sign-off, exclusion, unaffordable services or an equally accountable alternative with better outcomes would weaken it.
These are HBC’s judgments and questions for further research. They do not show actual benefits or costs, establish workers’ preferences or authorise us to speak for affected people.
Three questions about the effects
- People’s say and decision-making power
- Can affected people get useful information, make choices, challenge a failure or help shape a collective decision?
- Income and economic power
- Who receives pay or bargaining power, who bears costs and who is excluded?
- Access to human services
- Can people who need a human obtain one, including those outside a service or contract?
We have not evaluated results across all three aims for any example. A written duty, payment or right to review does not show whether these aims are achieved.
Questions that need more research
- Trace real assignments from AI input to the professional’s reasons, ability to reject output and information given to clients.
- Check later territorial versions, reported disciplinary decisions, available remedies and how predominant intellectual work is interpreted.
- Compare costs, entry-level learning, pay and access with more specific human-decision duties and credible AI-assisted alternatives.
Sources and versions checked
Follow the links to the passages listed below. Sources can change. The date we checked a source does not confirm every later change or whether the rule applies to you.
- IT132
Italy: Law 132/2025, Article 13 ↗
- Where to look
- Article 13; printed p 5, PDF 7
- Version checked
- 25 September 2025 enacted publication; inspected 15 November 2026
- ITTR26
Terni engineers’ code ↗
- Where to look
- Cover; Articles 2,13,23–24, pp 1,4,10,19
- Version checked
- 23 June 2026 text; Terni ratification 16 September 2026