欧盟理事会禁止AI裸体化工具，纽约强制要求水务公司网络安全，波兰核研究中心遭受网络攻击

← 返回博客

---

🇪🇺 European Council Adopts AI Act Amendments — Nudification Tools Banned, Biometric Rules Tightened

The European Council on Friday released its proposal for streamlining the EU's landmark AI Act, adding significant new provisions that go beyond the European Commission's original amendment package. The Council's position includes a new prohibition on AI practices related to generating non-consensual sexual and intimate content, as well as child sexual abuse material — a direct response to the Grok deepfake scandal that erupted in late 2025.

The move comes just days after the European Parliament's own committee greenlit a similar ban, suggesting the final negotiated text will almost certainly include the measure. The European Commission had launched a formal investigation into X and its Grok feature in January 2026 after the chatbot was used to generate millions of non-consensual intimate images.

Beyond the nudification ban, the Council's proposal reinstates strict necessity requirements for processing special categories of personal data — including biometric data — for bias detection and correction purposes. This is a significant departure from the Commission's original proposal, which had loosened these requirements in the name of "simplification."

What This Means for AI Developers and Deployers

• Prohibited practices list expands. Any AI system capable of generating non-consensual intimate imagery will be classified as a prohibited practice under the AI Act. This affects not just dedicated deepfake tools but any generative AI system that lacks adequate safeguards — including general-purpose models with image generation capabilities.
• High-risk AI database registration maintained. The Council rejected the Commission's proposal to relax database registration requirements. Providers who claim their AI systems are not high-risk must still register them in the EU database, maintaining transparency and accountability.
• Biometric data under strict scrutiny. The reinstatement of "strict necessity" for processing biometric data means organizations using AI for facial recognition, emotion detection, or behavioral biometrics must demonstrate that no less intrusive alternative exists. This aligns with GDPR's existing data protection framework and the European Data Protection Board's guidance.
• Timeline extensions create breathing room. The Commission's proposal to extend high-risk AI system deadlines by up to 16 months, combined with expanded SME exemptions, gives organizations more time to prepare — but the nudification ban and biometric rules take effect immediately upon final adoption.

---

🏛️ New York Mandates Cybersecurity for Water Utilities — America's NIS2 Moment

New York has become the first US state to impose mandatory cybersecurity regulations on water and wastewater utilities, creating requirements that closely mirror the EU's NIS2 Directive approach to critical infrastructure protection. The regulations, proposed in July 2025 and recently approved, take effect with a compliance deadline of 2027.

The rules require community water systems serving more than 3,300 people to implement cybersecurity training, incident response plans, reporting requirements, and designate a cyber lead. Larger systems serving over 50,000 people face additional obligations. New York has backed the mandate with a $2.5 million grant program to help underfunded utilities meet the new standards.

"We could not wait for stalled federal mandates while cyber threats intensify," said Michaela Lee, New York's acting chief cyber officer, referencing China's Volt Typhoon campaign that has been pre-positioning within US water infrastructure for potential destructive action.

The NIS2 Parallel: Why Europeans Should Pay Attention

Requirement	New York Water Regs	NIS2 Directive
Scope	Water systems >3,300 people	Essential entities including water supply
Incident Response	Mandatory response & recovery plans	Incident handling & business continuity (Art. 21)
Reporting	Mandatory incident reporting	24h early warning, 72h notification (Art. 23)
Training	Certified operator cyber training	Cybersecurity awareness training (Art. 21)
Leadership	Designated cyber lead	Management body accountability (Art. 20)
Funding Support	$2.5M grant program	No direct EU funding mechanism

The convergence between US state-level and EU regulatory approaches is striking. Both recognize that critical infrastructure operators — particularly smaller ones — lack the resources and expertise to implement cybersecurity on their own. The key difference: New York provides direct financial support, while NIS2 relies on enforcement pressure without a corresponding funding mechanism.

• State action fills the federal vacuum. After water industry lobbyists successfully fought off EPA cybersecurity mandates in 2024, states are stepping in. New York's approach — starting with financial services and healthcare before moving to water — mirrors NIS2's sector-by-sector implementation strategy.
• Volt Typhoon drives urgency. Both New York regulators and European authorities have cited Chinese state-sponsored pre-positioning in critical infrastructure as the primary threat driver. This shared threat landscape makes transatlantic regulatory alignment increasingly likely.
• Grant programs acknowledge reality. New York's $50,000 assessment grants and $100,000 upgrade grants acknowledge what NIS2 implementation has also revealed: many essential service operators simply cannot afford cybersecurity compliance without external support.

---

☢️ Poland's Nuclear Research Centre Repels Cyberattack — Suspected Iranian Origin

Poland's National Centre for Nuclear Research (NCBJ) — the country's principal nuclear research institution and operator of the MARIA research reactor — disclosed this week that hackers targeted its IT infrastructure. The institute stated that its security systems and incident response procedures detected and blocked the attack before any compromise occurred.

While NCBJ did not attribute the attack, Reuters reported that Polish authorities found indicators suggesting Iranian involvement, though investigators caution these could be false flags. The timing is notable: Poland's Defense Minister recently stated that Poland is not participating in the current Middle East conflict.

This incident follows a pattern of escalating cyber operations against Polish critical infrastructure. In January 2026, Russia's APT44 (Sandworm) attacked Poland's power grid, impacting 30 distributed energy facilities. An ICCT report from February placed Poland among the top targets of Russian cyber actors, with 31 confirmed incidents between mid-2025 and early 2026.

NIS2 and Nuclear Security Implications

• NIS2 incident reporting triggered. Even though the attack was blocked, NIS2's reporting requirements apply to "near-miss" incidents that could have had significant impact on essential services. A cyberattack on nuclear research infrastructure — even unsuccessful — falls squarely within the scope of incidents requiring notification to national CSIRTs and competent authorities.
• Successful defense validates NIS2's approach. NCBJ's statement that its security systems and procedures worked exactly as designed is a textbook example of what NIS2 aims to achieve: organizations with mature incident detection and response capabilities that can stop threats before they cause damage. The 24-hour early warning requirement exists precisely to share this threat intelligence with peer organizations.
• Nuclear sector faces elevated threat. The convergence of Iranian, Russian, and Chinese cyber operations targeting European critical infrastructure makes the case for NIS2's stringent requirements for essential entities. Nuclear research facilities handle dual-use technologies and sensitive scientific data that are high-value targets for state-sponsored espionage.
• Cross-sector threat intelligence sharing is essential. Poland's experience — attacked by Russian APTs in January and now targeted by suspected Iranian actors in March — demonstrates why NIS2 mandates information sharing between essential entities and national authorities. No single organization can maintain awareness of threats from multiple state-sponsored adversaries.

---

🇺🇸 US Sanctions North Korean IT Worker Networks — Compliance Implications for Global Employers

The US Treasury Department has sanctioned six individuals and two companies supporting North Korea's massive IT worker fraud scheme, targeting operations in Vietnam, Laos, and Spain. The sanctioned entities include Amnokgang Technology Development Company (a North Korean front company managing worker delegations) and Quangvietdnbg International Services Company (a Vietnamese firm that converted approximately $2.5 million for North Korean actors between 2023 and 2025).

The scheme involves thousands of North Korean citizens secretly obtaining employment at US and European companies under false identities, earning high-paying technology salaries that are funneled back to Pyongyang. The operation generated nearly $800 million in 2024 alone, according to US officials.

What This Means for European Employers

• Know-Your-Employee due diligence is critical. European companies hiring remote IT workers — particularly freelancers and contractors — must implement robust identity verification processes. North Korean operatives use sophisticated forged credentials, AI-generated profiles, and proxy identities to pass background checks.
• OFAC sanctions compliance extends globally. Any European company doing business with sanctioned entities — even unknowingly — risks secondary sanctions. The designation of a Spanish national in this round demonstrates that the scheme extends well beyond Asia.
• NIS2 supply chain provisions apply. Organizations outsourcing IT work to freelancers or staffing agencies must assess the security of their supply chain, including the risk of inadvertently employing sanctioned individuals. NIS2 Article 21(2)(d) explicitly requires supply chain security measures.
• GDPR employment data obligations. Employers who discover they've unknowingly hired a North Korean operative face a complex GDPR situation: the personal data collected during the employment relationship was obtained through fraud, potentially triggering breach notification requirements and creating data retention questions.

---

📡 Regulation Radar — Key Dates & Deadlines

Date	Framework	Milestone
Q2 2026	EU AI Act	Council-Parliament trilogue negotiations on AI Act amendments (nudification ban, biometric rules)
Aug 2, 2026	EU AI Act	Obligations for high-risk AI systems take effect (potentially extended 16 months under amendment proposal)
2027	NY Water Regs	New York water utility cybersecurity regulations compliance deadline
Q3 2026	DORA	First critical third-party ICT provider designations by European Supervisory Authorities
Ongoing	OFAC	Active enforcement against employers of North Korean IT workers — continuous sanctions list monitoring required

---