Italy software supply chain and AI security rules: a 90-day plan

If you build, buy or run software in Italy, the rules that reach your code and your suppliers are already in force: Legislative Decree 138/2024, which transposes NIS2 and is enforced by the National Cybersecurity Agency (ACN); the National Cybersecurity Perimeter for the most critical operators; Law 90/2024 for public administrations; DORA for financial entities supervised by the Bank of Italy, Consob, IVASS and COVIP; the EU Cyber Resilience Act, whose reporting duty started on 11 September 2026; and Law 132/2025, Italy's national AI law, which sits beside the EU AI Act. The common answer to all of them is the same: know what is in your software, know who supplied it, and be able to show both within hours of an incident.

This guide takes each regime from its primary source, names the supervisor, lists what it asks of software and open source, and ends with a 30/60/90-day plan that works with any tooling. It is a practical reading of public texts as they stand in late October 2026, not legal advice.

Key takeaways

• NIS2 is law in Italy and the phase-in is over. Legislative Decree no. 138 of 4 September 2024 has been in force since 16 October 2024. ACN is the competent authority and CSIRT Italia receives notifications. Article 42 gave entities nine months from ACN's inclusion notice for incident duties and eighteen months for governance and security measures; for entities notified in spring 2025, both have run.
• Registration is annual and the incident clock is short. Entities register on ACN's platform between 1 January and 28 February each year (Article 7). Article 25 sets a 24-hour pre-notification, a 72-hour notification and a final report.
• The Perimeter is stricter and separate. Decree-Law 105/2019 requires Perimeter operators to keep ICT asset lists, notify incidents, apply prescribed measures and send certain ICT purchases to the national evaluation centre (CVCN) first.
• Public bodies have their own clock. Law 90/2024 requires many public administrations to report incidents within 24 hours and notify fully within 72 hours.
• DORA applies through four Italian supervisors: the Bank of Italy, Consob, IVASS and COVIP, under Legislative Decree 23/2025.
• Italy has a national AI law. Law 132/2025, in force since 10 October 2025, names AgID and ACN as national AI authorities, gives ACN supervisory powers over AI systems and calls cybersecurity an essential precondition across the AI life cycle.
• Safeguard capabilities covered here, all GA: SBOM generation and analysis, SCA and deep transitive dependency scanning, vulnerability management with EPSS and KEV, reachability analysis, policies and gates, the Compliance and GRC suite with framework pages for Italy ACN, NIS2, DORA and the EU CRA, TPRM SBOM requests, CSAF and VEX upload, Trust Center, AI-BOM discovery, AI Gateway, MCP Guardrail, agent identity, autonomous remediation (ships switched off), the local runner, Zero AI mode, and private VPC, on-prem and air-gapped deployment.

Who this applies to in Italy

Italy layers national security law on top of EU law, so one organisation can sit in two or three regimes at once, each with its own notification route. Start by finding your boxes.

NIS2 essential and important entities. Medium and large organisations in the sectors listed in the annexes to Legislative Decree 138/2024 fall in scope: energy, transport, banking, financial market infrastructure, health, water, digital infrastructure, ICT service management, space, postal services, waste, chemicals, food, manufacturing, digital providers and research, plus public administrations. Some categories, such as DNS providers and trust service providers, are in scope regardless of size. ACN is the competent authority, working with sector authorities.

Perimeter operators. A limited, non-public list of public and private bodies whose networks and systems support essential state functions sit inside the National Cybersecurity Perimeter (Perimetro di sicurezza nazionale cibernetica), run by ACN. Many are also NIS2 entities, and the Perimeter rules apply on top for the ICT assets they list.

Public administrations. Article 1 of Law 90/2024 names central public administrations, regions, the autonomous provinces of Trento and Bolzano, metropolitan cities, municipalities above 100,000 inhabitants and all regional capitals, larger public transport companies and local health authorities (aziende sanitarie locali), plus their in-house IT companies. Public bodies moving to the cloud also work under ACN's cloud rules.

Regulated finance. Banks, payment and e-money institutions, investment firms, asset managers, insurers and pension funds fall under DORA, Regulation (EU) 2022/2554, supervised in Italy by the Bank of Italy, Consob, IVASS and COVIP.

Software makers and suppliers. Any manufacturer placing products with digital elements on the EU market is covered by the Cyber Resilience Act, Regulation (EU) 2024/2847. And if you sell software or services to an Italian bank, utility, hospital or ministry, you will meet NIS2, DORA and Perimeter duties in contract clauses, questionnaires and audits even where the law does not name you.

Anyone processing personal data. The GDPR applies, enforced by the Garante per la protezione dei dati personali, one of Europe's most active regulators on AI services.

The rules, regulator by regulator

NIS2: Legislative Decree 138/2024 and ACN

Legislative Decree no. 138 of 4 September 2024 (Decreto legislativo 4 settembre 2024, n. 138) transposes Directive (EU) 2022/2555. It was published in Official Gazette no. 230 of 1 October 2024 and entered into force on 16 October 2024. Its parts that matter most for software teams are these.

Authority and CSIRT. ACN is the single national competent NIS authority and the single point of contact under Article 8 of the directive. CSIRT Italia, the national incident response team within ACN, receives incident notifications.

Registration (Article 7). From 1 January to 28 February of each year, entities that may be in scope register or update their registration on ACN's digital platform. They give their contacts, a designated point of contact and their sectors and entity types. By 31 March each year ACN draws up the list of essential and important entities and tells registered entities, through the platform, whether they are on it.

Governance (Article 23). Management bodies must approve the risk management measures, oversee their implementation and follow training, and they can be held responsible for breaches. This is why NIS2 is a board topic in Italy, not only an IT one.

Security measures (Article 24). Entities must adopt proportionate technical, operational and organisational measures following Article 21 of the directive, including supply chain security with direct suppliers, and security in acquisition, development and maintenance, including vulnerability handling and disclosure.

Incident notification (Article 25). Essential and important entities notify CSIRT Italia, without undue delay, of every incident with a significant impact on their services. An incident is significant if it has caused or can cause severe operational disruption or financial loss for the entity, or considerable material or non-material damage to others. The sequence is a pre-notification within 24 hours of becoming aware of the incident, saying where possible whether it may be malicious or have a cross-border impact; a notification within 72 hours with an initial assessment, severity, impact and indicators of compromise where available; intermediate reports on request; and a final report.

Sanctions (Article 38). For breaches of the core obligations, including risk management and incident notification, essential entities face fines of up to 10 million euros or 2% of total worldwide annual turnover, whichever is higher, and important entities up to 7 million euros or 1.4%. Lower caps (0.1% and 0.07% of turnover) apply to other breaches such as failing to register. Our related post covers the tiers in more detail: Italy's NIS2 decree and its tiered sanctions.

Transition (Article 42). Until 31 December 2025, entities had nine months from receiving ACN's inclusion notice to meet the incident notification duties of Article 25, and eighteen months to meet the duties in Articles 23, 24 and 29. ACN could set basic procedures and specifications to help entities comply in that period. For an entity notified in April 2025, the nine months ran out in January 2026 and the eighteen months in October 2026. If you are an Italian NIS2 entity, the security measures are not a future project any more.

ACN's basic specifications

ACN has used that power. Its determination no. 164179 of 14 April 2025 sets the basic specifications for the NIS obligations: baseline security measures for important and for essential entities, and the baseline significant incidents each must notify. Italy's communications regulator, AGCOM, builds on it in its July 2025 decision on security governance, which shows how far the determination now reaches into sector rules.

For software teams, the consequence is that supply chain measures in an essential entity become questionnaire items for every vendor in its chain. ACN's portal publishes the determination and its annexes; read them in full before you map controls, because this guide does not reproduce their numbering.

The National Cybersecurity Perimeter

The Perimeter was created by Decree-Law no. 105 of 21 September 2019, in force from 22 September 2019 and converted with amendments by Law no. 133 of 18 November 2019. Its implementing acts fill in the detail:

• DPCM no. 131 of 30 July 2020, in force from 5 November 2020, sets how bodies are identified for the Perimeter and how they draw up and transmit the list of ICT assets that support their essential functions or services.
• DPR no. 54 of 5 February 2021, in force from 8 May 2021, sets the procedure for notifying planned ICT purchases to the CVCN (Centro di Valutazione e Certificazione Nazionale), now within ACN, and the evaluation and testing that can follow.
• DPCM no. 81 of 14 April 2021, in force from 26 June 2021, sets the incident notification rules and the security measures. Article 3 requires Perimeter operators, from 1 January 2022, to notify CSIRT Italia of the incident types listed in its Annex A that affect their listed ICT assets.

In plain terms, a Perimeter operator must document every ICT asset behind an essential function, apply the prescribed measures, notify incidents to CSIRT Italia on short deadlines, and tell the CVCN before buying certain categories of ICT products so the CVCN can impose tests or conditions. If you supply software to a Perimeter operator, expect your product to appear in an asset list, possibly to go through CVCN testing, and to face questions about every component inside it.

Law 90/2024: public administrations and cyber procurement

Law no. 90 of 28 June 2024 (Legge 28 giugno 2024, n. 90) on strengthening national cybersecurity and cybercrime rules was published in Official Gazette no. 153 of 2 July 2024 and entered into force on 17 July 2024.

Article 1 requires the bodies listed above to report the incidents in ACN's taxonomy that affect their networks, information systems and IT services without delay and in any case within 24 hours of becoming aware of them, and to make a complete notification with all available information within 72 hours. The reports go through the procedures on ACN's website. The law also requires these bodies to have a cybersecurity structure and a named cybersecurity contact person (referente per la cybersicurezza).

Article 14 required a prime ministerial decree, proposed by ACN, setting for specific technology categories the essential cybersecurity elements that public buyers must consider when procuring ICT used in contexts linked to strategic national interests. If you sell software to Italian public bodies, read that decree as part of tender preparation, because it shapes what a buyer can score.

ACN's cloud rules for public administration

Since the 2021 national cloud strategy, Italian public bodies classify their data and services as ordinary, critical or strategic, and may use only cloud infrastructure and cloud services that ACN has qualified for the matching level. ACN's single regulation on digital infrastructure and cloud services for public administration has applied since 1 August 2024. It sets the qualification levels, the requirements for each and the procedure for obtaining and keeping a qualification, and ACN keeps a catalogue of qualified services that public buyers consult.

For software vendors, qualification attaches to the cloud service a public body consumes, so a SaaS product must be qualified at the level its data needs. The rules govern where and how a service runs; they do not replace vulnerability handling in the software inside it.

DORA and the Bank of Italy, Consob, IVASS and COVIP

DORA has applied across the EU since 17 January 2025. Legislative Decree no. 23 of 10 March 2025, published in Official Gazette no. 58 of 11 March 2025 and in force from 12 March 2025, adapts Italian law to it. Article 3 names the competent authorities under Article 46 of DORA: the Bank of Italy, Consob, IVASS and COVIP, each for the entities it already supervises. The Bank of Italy is also the authority for Cassa depositi e prestiti and, for the national provisions, for financial intermediaries and Bancoposta. The DORA articles that land on software supply chains are Article 8 (identify and document ICT assets and their dependencies), Article 9 (protection, including integrity and patching), Articles 17 to 19 (classify and report major ICT-related incidents), Article 28 (ICT third-party risk and the register of information) and Article 30 (key contractual provisions such as data location, security requirements, audit rights and exit strategies). For a software vendor to an Italian bank or insurer, they arrive as contract clauses, a register entry and a questionnaire, and recur at renewal.

The EU Cyber Resilience Act

The CRA, Regulation (EU) 2024/2847, entered into force on 10 December 2024. The Commission's page confirms that reporting obligations apply from 11 September 2026 and the main obligations from 11 December 2027. Under Article 14, manufacturers must report actively exploited vulnerabilities and severe incidents affecting their products with digital elements through the single reporting platform, starting with an early warning within 24 hours, with the CSIRT of their main establishment as coordinator. For manufacturers established in Italy that is CSIRT Italia. The essential requirements in Annex I, including the duty in Part II, point 1, to identify and document components by drawing up an SBOM covering at least the top-level dependencies, apply from December 2027. The Commission published practical guidance for manufacturers on 27 July 2026.

The EU AI Act and Italy's Law 132/2025

The EU AI Act, Regulation (EU) 2024/1689, applies in stages. Prohibited practices and AI literacy duties have applied since 2 February 2025, and the governance rules and obligations for general purpose AI models since 2 August 2025. The Commission's page says that, after the political agreement on the "AI Omnibus" simplification, the rules for high-risk use cases in Annex III apply from 2 December 2027 and those for high-risk AI embedded in regulated products under Annex I from 2 August 2028.

Italy's own AI law is Law no. 132 of 23 September 2025 (Legge 23 settembre 2025, n. 132, Disposizioni e deleghe al Governo in materia di intelligenza artificiale), published in Official Gazette no. 223 of 25 September 2025 and in force since 10 October 2025. Three provisions matter here.

• Article 3(5) says the law creates no new obligations beyond those in the AI Act for AI systems and general purpose AI models. It complements the Regulation rather than adding a second compliance regime.
• Article 3(6) says cybersecurity must be ensured, as an essential precondition, across the whole life cycle of AI systems and models.
• Article 20 designates AgID and ACN as the national AI authorities. AgID promotes AI and handles notification, assessment, accreditation and monitoring of conformity assessment bodies. ACN is responsible for supervision of AI systems, including inspections and sanctions, and for the cybersecurity side of AI. The Bank of Italy, Consob and IVASS keep their market surveillance role for AI in the financial sector, as Article 74(6) of the AI Act allows.

For a security team the message is simple: ACN, which already supervises your NIS2 compliance, can also inspect your AI systems.

The Garante and the GDPR

The Garante enforces the GDPR in Italy: Article 32 on security of processing, Article 33 on breach notification to the authority within 72 hours where feasible, and Article 34 on informing people at high risk.

What the rules ask of software and open source

Each regime uses its own words, but an auditor or supervisor's questions fall into the same groups. This table maps them to the evidence you should be able to hand over.

Almost every line needs a current, versioned list of what is inside your software and who supplied it. That is why the SBOM keeps appearing even where Italian texts never use the word.

Real public incidents in or affecting Italy

These are public events from official or primary sources in the last three years. They are listed to show what the rules respond to, not to judge the organisations involved.

• A backdoor in a widely used open source library, March 2024. On 29 March 2024 CISA reported that malicious code had been embedded in XZ Utils versions 5.6.0 and 5.6.1, a data compression library present in Linux distributions, assigned CVE-2024-3094. CISA advised developers and users to downgrade to an uncompromised version such as 5.4.6 and to hunt for malicious activity. The first question every team, in Italy as elsewhere, had to answer was whether those versions were anywhere in its estate.
• The Garante's decision on an AI chatbot, December 2024, and the court ruling that followed. On 20 December 2024 the Garante announced that it had closed its investigation into OpenAI's ChatGPT service, imposing a 15 million euro fine and a six-month public information campaign. Its findings included that the company had not notified the authority of a personal data breach it suffered in March 2023. The Garante's page now notes that its decision no. 755 of 2 November 2024 has been temporarily removed after the Court of Rome, in judgment no. 4153/2026 published on 18 March 2026, upheld the company's challenge.
• The Garante's order against DeepSeek, January 2025. On 30 January 2025 the Garante ordered, urgently and with immediate effect, a limitation on the processing of Italian users' data by the two companies behind the DeepSeek chatbot, after they replied that they did not operate in Italy and that European rules did not apply to them. It opened an investigation at the same time.

The personas who own the problem

The CISO. Owns the answer to ACN, the board and the sector supervisor. Needs one current view of which systems are in scope, which suppliers matter and whether the 24-hour clock can be met. Deadlines: the annual registration window, ACN or Bank of Italy requests, and any incident. Under Article 23 the management body approves the measures and can be held responsible, so the CISO also needs a board report a non-specialist can sign.

The head of AppSec or VP Engineering. Owns the SBOMs, the vulnerability backlog and the release gates. Needs SBOMs on every release, a way to decide which vulnerabilities matter first and proof that fixes ship. Their trigger is a customer questionnaire, a CSIRT Italia alert, an open source event like XZ Utils or, for manufacturers, the CRA reporting duty.

The GRC or compliance lead, and the cybersecurity contact person. Owns the ACN registration, the mapping of ACN's basic specifications to evidence and, in public bodies, the role of referente per la cybersicurezza under Law 90/2024. Needs evidence by clause and an audit trail before an inspection or the annual registration update.

Procurement and vendor risk. Owns supplier assessments, the DORA register of information and, for Perimeter operators, CVCN notifications. Needs SBOMs and security statements from vendors and a record of what each sent and when, usually at renewal or tender time.

The DPO. Owns GDPR breach notification and the data protection side of AI use. Needs to know within hours whether an incident involves personal data, and which AI services staff use.

The head of digital transformation in a public body. Owns cloud migration under ACN's rules and often the AI adoption plan. Needs to know which services are qualified and what suppliers can show about the components inside.

A practical 30/60/90-day plan

This plan works with any tooling. Reorder it around whichever deadline is closest for you: an ACN inspection, a DORA register submission, a tender, or the next registration window in January.

Days 1 to 30: scope, owners and inventory

1. Confirm your status in writing. Essential or important entity, Perimeter operator, public administration under Law 90/2024, DORA financial entity, CRA manufacturer, or a supplier to any of these. Note the source of each answer.
2. Name the owners. The CISO, the NIS point of contact registered with ACN, the cybersecurity contact person where Law 90/2024 applies, the DPO, and a technical lead per product or critical system.
3. List essential services and the systems behind them. This is the Perimeter asset list, the base of your NIS2 risk analysis and the DORA Article 8 inventory.
4. Generate an SBOM for every product and critical system. SPDX or CycloneDX, with transitive dependencies, versions, suppliers and hashes where you can, stored with the release it describes.
5. List your ICT suppliers. Reconcile against the DORA register if you have one. Mark suppliers whose software runs inside critical systems.
6. Write your notification map on one page. CSIRT Italia under NIS2, ACN under Law 90/2024, your financial supervisor under DORA, the Garante under the GDPR, the single reporting platform under the CRA, each with its clock.

Template for the asset and SBOM register:

Days 31 to 60: vulnerability handling, incidents and suppliers

1. Match components against vulnerability data. Use CVE records plus EPSS and the CISA Known Exploited Vulnerabilities catalogue, then check whether the vulnerable code is reachable in your application.
2. Set fix targets by severity and exposure. Write them into a vulnerability management policy an inspector can read, including how you handle a component with no fix yet.
3. Test the 24-hour clock. Run a tabletop in which a library in a critical system is reported as actively exploited, as XZ Utils was. Time three answers: which systems and versions are affected, is it significant under Article 25, and who must be told by when.
4. Publish a disclosure route. A security contact and a coordinated vulnerability disclosure policy; the CRA expects manufacturers to have one.
5. Send supplier requests using the checklist below, with a deadline and a reminder.
6. Add release gates. Stop releases that carry known exploited vulnerabilities unless an exception with a justification, an owner and an expiry date is approved.

Supplier request checklist:

Days 61 to 90: evidence, AI and where your tools run

1. Map each measure to evidence. For NIS2, against ACN's basic specifications for your category; for DORA, against Articles 5 to 9, 17 to 19, 28 and 30. Record where each piece of evidence lives and who refreshes it.
2. Inventory AI use. List AI models, services and agents across code, cloud and SaaS, with owners, purpose and the access each holds. This is the base for AI Act classification and any ACN question under Law 132/2025.
3. Decide where your security tools run. If source code or data cannot leave your network, write that into your tool selection criteria before you buy.
4. Prepare a customer pack. An SBOM per supported version, your disclosure policy, VEX statements for headline vulnerabilities and a short security overview.
5. Report to the board. One page: scope, gaps, dates, owners and the decisions you need.

Common mistakes

• Treating registration as the job. Registering on ACN's platform starts the obligations. The measures, the evidence and the incident capability are the work.
• Mixing up the clocks. NIS2, the Perimeter, Law 90/2024, DORA, the CRA and the GDPR each have their own notification route and timing. Keep them on one page and rehearse them.
• Producing SBOMs once. An SBOM from last quarter cannot answer this morning's CSIRT Italia alert. Generate them on every release and keep the history.
• Forgetting the supply chain runs both ways. As a supplier you will be asked for evidence; as a customer you must ask for it.
• Leaving AI out of the inventory. AI services and coding assistants arrive through SaaS sign-ups and developer tools. The Garante has shown it will act, and ACN now supervises AI systems.
• Mapping to the wrong numbering. When you write to ACN, use the decree's articles and the identifiers in ACN's determination, not a vendor's own control IDs.

How Safeguard does this

Everything in the plan above works without Safeguard. This section shows how the same steps run in Safeguard, using only what Safeguard's documentation describes. Every capability named here is generally available (GA).

1. Turn on the Italian and EU framework pages. In the Compliance and GRC suite (GA), the Italy ACN Cybersecurity, NIS2, DORA and EU Cyber Resilience Act frameworks are on in a workspace unless you switch them off under Compliance › Manage. Each page lists the requirements a scan decides, plus catalogue controls you evidence with tests, uploads, policies or attestations. The identifiers on those pages are Safeguard's own labels, not ACN's or the EU's numbering, so map them to the legal text before you quote them to an authority.
2. Generate SBOMs on every scan. SBOM generation and analysis (GA) produces SPDX and CycloneDX, and SCA with deep transitive dependency scanning (GA) builds the full component inventory. On the Italy ACN page, each scan is checked for a populated component inventory (the asset list analogue), known suppliers, integrity hashes, recorded versions for vulnerability management, components identifiable fast enough to scope an incident, and dated documentation. The NIS2 and DORA pages run similar checks for supply chain security, incident readiness and asset records.
3. Scan without sending source code. The local runner (GA) is part of the Safeguard CLI. As the docs describe it, you install the CLI with the one-line installer, run safeguard, type /login, approve the session in your browser, and the machine starts polling the platform for work. It runs each scan job locally and uploads only the resulting SBOM and findings, never your source code, and it works on networks where only outbound HTTPS is allowed. For headless use the docs show:

# Start without the terminal UI and serve assistant turns for this workspace
safeguard runner start --workspace ./path/to/repo
# Is it running, and what is it doing
safeguard runner status
# Drain in-flight work and stop
safeguard runner stop

4. Prioritise what is exploited and reachable. Vulnerability management (GA) matches components against advisories and adds EPSS and CISA KEV context. Reachability analysis (GA) ranks the vulnerable code paths your application actually calls.
5. Gate releases and record exceptions. Policies and gates (GA) are set under Settings › Policies with a policy builder. Each policy's action is Block (the gate fails), Warn (the gate passes with a warning naming the policy) or Notify (the gate passes and a notification is sent). When a legitimate exception is needed, a developer requests one with a business justification and an expiration date, approvers review and decide, and the audit trail is kept. With policy outcomes on for your workspace, the docs show a CI job failing when a gate fails:

safeguard ci scan --wait --fail-on-gate <gate>

6. Collect evidence from suppliers. TPRM (GA) keeps the vendor directory. Its SBOM requests let you send single or bulk requests from a template that sets SPDX or CycloneDX format, required fields, submission instructions and a deadline. Status runs from Sent to Submitted and then Validated or Rejected, with automatic reminders before and after the deadline. That gives procurement the trail behind a DORA register entry or a NIS2 supply chain measure.
7. State what is and is not affected. CSAF and VEX upload (GA) holds exploitability statements you can share with customers, and helps when a public buyer or a bank asks whether a published vulnerability affects your product.
8. Fix, with a human in charge. Autonomous remediation (GA) can open tested, policy gated fix pull requests, singly or in bulk across projects. It ships switched off, with every category set to Manual, so your team decides what runs on its own.
9. Inventory and govern AI. AI-BOM discovery (GA) inventories AI assets, the starting point for AI Act classification. AI Gateway (GA) is an LLM firewall that applies guardrail policies for prompt injection, jailbreak attempts and PII egress to model traffic you route through it. MCP Guardrail (GA) generates guardrail proxy URLs for MCP servers, and agent identity (GA) tracks which agent holds what access, at what level.
10. Share posture with buyers. Trust Center (GA) publishes your posture and documents such as SBOMs and VEX to customers and helps answer security questionnaires.
11. Choose where it runs. Safeguard's SaaS is multi-region across US, EU and Gov regions; no Italy-specific region is offered or implied. Safeguard's EU region is in Frankfurt, Germany, so data stays in the EU. For residency, the GA options are private VPC or dedicated tenancy, full on-prem, and fully air-gapped with the AI models running inside your environment and signed offline snapshot delivery. Zero AI mode (GA) runs the whole platform deterministically with no LLM in the loop, and bring your own model (GA) is available for teams with an approved model list. Safeguard states that customer code, SBOMs and telemetry are not used to train shared models.

Safeguard's posture wording is limited to what its Trust Center says: SOC 2 Type II audit in progress, and designed for FedRAMP High. Safeguard does not claim any Italian qualification or assessment, including ACN cloud qualification or a CVCN evaluation.

What Safeguard does not do

• It does not make you compliant. It helps you produce and keep the evidence. Deciding that an incident is significant, a supplier acceptable or a measure adequate stays with you and your authorities.
• It does not notify anyone for you. It helps you answer "which products and versions are affected" quickly. Filing with CSIRT Italia, ACN, your financial supervisor, the CRA platform or the Garante is your team's job.
• It is not an ACN-qualified cloud service for public administration. Where a body's data needs a qualified service, Safeguard's SaaS does not satisfy that. The on-prem, air-gapped and local runner options let the body run the software inside an environment it already has approved.
• It does not perform CVCN evaluations or replace the Perimeter procedure.
• It does not write contracts or legal analysis. DORA Article 30 clauses and NIS2 applicability decisions need your legal team.
• It does not fix vendor appliances or hosted services you buy. That fix comes from the vendor; Safeguard helps you track the request and the evidence.
• It does not replace incident response. Containment, forensics and communication stay with your processes.

FAQ

Is NIS2 in force in Italy?

Yes. Legislative Decree 138/2024 has been in force since 16 October 2024, and its transitional periods have now run for entities notified in 2025.

Who supervises NIS2 in Italy?

The National Cybersecurity Agency (Agenzia per la cybersicurezza nazionale, ACN), as single competent authority and point of contact. CSIRT Italia, within ACN, receives incident notifications.

When must an Italian NIS2 entity report an incident?

For a significant incident, Article 25 requires a pre-notification to CSIRT Italia within 24 hours, a notification within 72 hours and a final report. Other regimes may add their own reports.

When is the NIS2 registration window in Italy?

From 1 January to 28 February every year, on ACN's platform, under Article 7.

What is the National Cybersecurity Perimeter?

A regime created by Decree-Law 105/2019 for bodies whose systems support essential state functions. Operators keep ICT asset lists, apply prescribed measures, notify incidents under DPCM 81/2021 and notify certain ICT purchases to the CVCN under DPR 54/2021.

Does DORA apply to Italian banks and insurers?

Yes, since 17 January 2025, supervised by the Bank of Italy, Consob, IVASS and COVIP under Legislative Decree 23/2025.

What does Italy's AI law require of security teams?

Law 132/2025 adds no obligations beyond the EU AI Act, but it makes ACN a supervisory authority for AI systems and calls cybersecurity an essential precondition across the AI life cycle. Keep an AI inventory and secure models and agents like other software.

Does a SaaS vendor selling to Italian public bodies need ACN cloud qualification?

If the body's data falls under ACN's cloud rules, the cloud service it buys must be qualified at the matching level. Check ACN's catalogue and the tender documents.

Is an SBOM legally required in Italy?

The CRA requires one from 11 December 2027 for products with digital elements. Until then, the Perimeter asset list, NIS2 asset and supply chain measures and DORA Article 8 all need the information an SBOM provides, which is why Italian buyers already ask for it.

Italian rules expect you to know your components and act fast when one is exposed. The section below explains how Safeguard helps close the gap between a published fix, or a missing one, and the software you actually run.

Fix it without waiting for upstream

Whatever the vulnerability, most of the risk sits in two gaps. The first is the time between a fix being published and that fix actually running in your production builds. The second, and worse, is the time before any fix exists at all, while a maintainer works out a patch and a release schedule. Waiting through both gaps is a habit, not a requirement. Safeguard gives teams three ways to close them.

1. Autonomous remediation: pull requests, not tickets

When a fixed version exists, Safeguard's Griffin models open a tested, policy gated pull request that moves the affected dependency to the safe version. You can run it on one repository or in bulk across every project that carries the vulnerable version. Remediation categories include dependencies, licenses, security posture and findings, and each one is set separately. Reachability analysis ranks the vulnerable code paths your application actually calls first, so the first pull requests you merge are the ones that matter most. Autonomous remediation ships switched off, with every category set to Manual, so your team decides how much runs on its own and what still needs a human approval.

2. Gold Open Source: the safe version in seconds, free

gold.safeguard.sh is free and needs no login. Search any CVE or package to see affected and fixed versions, CVSS, EPSS, CISA KEV status and a recommended safe version, or use the "Am I affected?" checker. The same data is available through a free read API, README status badges and RSS threat feeds.

3. Gold Registry: start from components that are already fixed

Gold Registry provides hardened, drop in replacements for upstream packages and container images. Every Gold artifact is held continuously to zero exploitable critical or high CVEs and a completed license check, with signed build provenance (signed with Safeguard's key and verifiable with cosign, SLSA Build Level 3 as the design target) and a public evidence page you can hand to an auditor. An artifact loses its Gold badge if it regresses. Teams can point their own package directory at the registry so hardened equivalents resolve without changing how developers install anything.

When upstream has no fix yet, or will not ship one, Safeguard first works to get the fix accepted upstream. If the issue is severe or upstream will not move, Safeguard builds a fork, fixes the vulnerability there and publishes a hardened Gold build, so you are not stuck on a maintainer's release schedule. You can also request a hardened, compatibility tested build of a specific package. Ask which packages it covers before you plan around it.

One honest limit: when the vulnerable code lives in a vendor appliance, a hosted service or an operating system runtime you do not build yourself, the fix still has to come from that vendor or distribution. The steps above apply to the open source dependencies and container images you build and ship, which is where most teams carry the largest and least tracked share of their exposure.

See how Safeguard would help

The first demo is a real CLI session. You install the CLI, run safeguard to open the interactive terminal and ask it to "fix the lodash CVE in this repo". It reads the finding and the advisory, asks before any command that changes something, edits package.json to lodash 4.17.21, runs npm install and npm test with your approval, prints a summary and pushes nothing. It then shows /permissions, /model and /scan. The same flow is how an Italian software house closes a vulnerability it may have to report under the CRA, with a human approving each change.

Click through it at your own pace, or open it full screen.

The second demo sets up the local runner on your own machine, approves it with its device code, then scans from the terminal. This is the option for Perimeter operators, public administrations and banks that will not let source code leave their network.

Click through it at your own pace, or open it full screen.

The third demo connects Claude to the Safeguard MCP server with OAuth and chooses which tools Claude can use. With ACN now supervising AI systems under Law 132/2025, it shows what scoped, reviewable agent access looks like in practice.

Click through it at your own pace, or open it full screen.

Quick check

Sources

• Normattiva: Legislative Decree 4 September 2024, no. 138 (NIS2 transposition)
• Decree 138/2024, Article 7 (registration)
• Decree 138/2024, Article 25 (incident notification)
• Decree 138/2024, Article 38 (sanctions)
• Decree 138/2024, Article 42 (first application)
• AGCOM: Decision 204/25/CONS, referring to ACN determination no. 164179 of 14 April 2025 (PDF)
• Normattiva: Decree-Law 21 September 2019, no. 105 (National Cybersecurity Perimeter)
• Normattiva: DPCM 30 July 2020, no. 131
• Normattiva: DPR 5 February 2021, no. 54
• Normattiva: DPCM 14 April 2021, no. 81, Article 3
• Normattiva: Law 28 June 2024, no. 90
• Law 90/2024, Article 14 (ICT procurement)
• Normattiva: Legislative Decree 10 March 2025, no. 23 (DORA adaptation)
• Decree 23/2025, Article 3 (competent authorities)
• Normattiva: Law 23 September 2025, no. 132 (artificial intelligence)
• Law 132/2025, Article 3
• Law 132/2025, Article 20 (national AI authorities)
• European Commission: Cyber Resilience Act
• European Commission: AI Act regulatory framework and timeline
• CISA: Reported supply chain compromise affecting XZ Utils (CVE-2024-3094)
• Garante: press release on the ChatGPT investigation and the Court of Rome note
• Garante: press release limiting DeepSeek's processing of Italian users' data
• Safeguard docs: Italy ACN Cybersecurity framework
• Safeguard docs: EU NIS2 framework
• Safeguard docs: EU DORA framework
• Safeguard docs: EU Cyber Resilience Act framework
• Safeguard docs: Local Runner
• Safeguard docs: Policies and Gates
• Safeguard docs: TPRM SBOM Requests

Related reading: the EU NIS2 directive and the software supply chain, DORA and the software supply chain in financial services, SBOM obligations under the EU CRA and the same plan for France.

To try the flows above, download Safeguard and read the Safeguard documentation.

Last reviewed: 25 October 2026. Product capabilities change; check docs.safeguard.sh for the current state.