July 2026 · 6 min read
EU AI Act Omnibus Final
Only 22 Days to Article 50 Enforcement
On July 9, 2026, the Council of the EU gave its final green light to the Digital Omnibus on AI, the most significant amendment to the AI Act since it entered into force. The amendment takes effect three days after publication in the Official Journal. Three weeks after that, on August 2, Article 50 transparency obligations become enforceable.
Key Definitions
EU AI Act Omnibus Final On July 9, 2026, the Council of the EU gave its final green light to the Digital Omnibus on AI, the most significant amendment to the AI Act since it entered into force. The amendment takes effect three days after publication in the Official Journal. Three weeks after that, on August 2, Article 50 transparency obligations become enforceable.
22 days. That is the window every organization providing or deploying AI systems to EU users has to complete Article 50 compliance.
Penalties: up to €35M or 7% of global annual turnover, whichever is higher. There is no transition period and no exemption window.
Article 50: What You Need to Know
Article 50 is the AI Act's core transparency provision. It does not distinguish between high-risk and low-risk. Any AI system in specific interaction scenarios must meet transparency requirements. Specifically:
Article 50(1): Interaction Disclosure
When users interact with an AI system (including chatbots, virtual assistants), they must be clearly informed that they are interacting with AI, unless it is obvious from the context and manner of interaction.
Article 50(2): AI-Generated Content Marking
AI-generated text, image, audio, or video content must be marked in a machine-readable format as AI-generated. This is the core control mechanism for synthetic content and deepfakes.
Article 50(3): Emotion Recognition and Biometric Classification
When using AI for emotion recognition or biometric-based classification, affected individuals must be informed.
Article 50(4): Deepfake Disclosure
AI-generated or manipulated images, audio, or video resembling real persons, objects, places, or entities (deepfakes) must be disclosed as artificially generated or manipulated.
Article 50(5): High-Risk Testing Environments
In real-world testing environments for high-risk AI systems, affected individuals must be informed.
The Omnibus grants a 4-month transition period (until December 2, 2026) for the Article 50(2) machine-readable marking requirement, and only for AI systems already placed on the market before August 2. All other Article 50 obligations (interaction disclosure, deepfake disclosure, etc.) remain on the original schedule.
What the Omnibus Actually Changed
The Digital Omnibus on AI is a targeted amendment that adjusts four areas:
1. High-Risk AI Obligations Deferred
Annex III high-risk AI obligations move from August 2, 2026 to December 2, 2027. Annex I embedded product AI moves to August 2, 2028. But Article 50 transparency obligations are untouched.
2. New Prohibitions
Article 5 adds a ban on AI-generated non-consensual intimate imagery ("nudifiers") and CSAM. Providers must assess foreseeable misuse risks and implement technical safeguards.
3. Watermarking Grace Period
Article 50(2) machine-readable marking delayed to December 2, 2026 for systems placed on the market before August 2, 2026.
4. Duplicative Requirements Reduced
AI systems already covered by other EU regulations (medical device regulation, machinery regulation) get harmonized compliance requirements.
High-risk deferred does not mean transparency deferred. Most headlines focus on the "deadline extension," which is good news for enterprises, but Article 50 was never delayed. In 22 days, AI systems still need to comply with the transparency rules.
22-Day Countdown: Who Needs to Act Now?
According to Responsible AI Labs' April 2026 survey, 78% of organizations have not taken meaningful steps toward compliance. Excluding organizations with no EU user exposure, three groups face the highest urgency:
- SaaS and software products serving EU users: If your chatbot, customer service AI, or content generation tool serves EU users, it must disclose AI interaction starting August 2. Interaction disclosure + content marking are the most immediate requirements.
- EU enterprises deploying AI agents: Internal agents that don't face end users may be partially exempt, but agent interactions with employees or partners may still trigger Article 50(1) obligations.
- Globally deployed AI platforms: If you operate in 50+ countries, the EU is just one jurisdiction. But the AI Act's penalty is based on global annual turnover, not just EU revenue.
What the AI Office Can Do After August 2
August 2 is both a compliance deadline and the date the AI Office begins full enforcement. From this date, the AI Office has the authority to:
- Request information and documentation from providers and deployers
- Require corrective actions, including modification or withdrawal of AI systems
- Initiate market surveillance procedures and impose penalties
- Issue compliance guidance and enforcement policies
Starting August 2, the AI Office can tell you that you are non-compliant, demand you stop using an AI system, and fine you up to €35M or 7% of global annual turnover.
Enforcement begins on August 2 and stays in force.
What Can You Do in 22 Days?
Building a complete AI governance system takes time. Three things must be done in 22 days:
1. Inventory: Identify all AI systems serving EU users
You don't need to govern everything at once, but you need to know which AI systems fall under Article 50. Agent Registry is step one.
2. Classify: Determine each system's transparency obligations
Which Article 50 obligations apply to each system? Check interaction disclosure, content marking and deepfake disclosure, then map them one by one.
3. Embed: Build disclosure and marking into runtime
Article 50 lands in runtime behavior, not in a compliance document. AI systems must automatically inform users during interaction and automatically embed markers in generated content. Policy documents cannot solve this. It requires a governance layer operating at runtime.
Steps 1 and 2 can be done manually (though not recommended). Step 3 must be embedded in the technical architecture of your AI systems. Without a governance layer, organizations must manually add disclosure and marking to every AI system, a task virtually impossible in 22 days.
Why Article 50's Date Never Moved
The EU AI Act took 3 years to legislate. The Omnibus took 8 months. Article 50's effective date never wavered, because legislators treat transparency as the foundation layer of AI governance.
Without transparency there is no auditability, and without auditability there is no governance. Without governance, AI is an uncontrollable cost.
August 2 is when the obligations begin.
FAQ
What do you need to know about Article 50?+
Article 50 is the AI Act's core transparency provision. It does not distinguish between high-risk and low-risk. Any AI system in specific interaction scenarios must meet transparency requirements. Specifically:
What did the Omnibus actually change?+
The Digital Omnibus on AI is a targeted amendment that adjusts four areas:
22-Day Countdown: Who Needs to Act Now?+
According to Responsible AI Labs' April 2026 survey, 78% of organizations have not taken meaningful steps toward compliance. Excluding organizations with no EU user exposure, three groups face the highest urgency:
Why is the AI Office's enforcement authority an often underestimated risk?+
August 2 is both a compliance deadline and the date the AI Office begins full enforcement. From this date, the AI Office has the authority to:
What Can You Do in 22 Days?+
Building a complete AI governance system takes time. Three things must be done in 22 days:
Related Articles
Agent traceability is becoming a bid condition
A new US bill directs NIST to write agent security standards in a year. Voluntary for most, mandatory for federal contractors.
EU designates ChatGPT a 'very large' search engine
ChatGPT is the first AI chatbot named a DSA 'very large search engine': 159M EU users, annual risk audits, algorithm transparency, ad library by Jan 2027.
China's AI Agent Recall Law Takes Effect July 15
On July 15, 2026, China's Implementation Opinions on AI Agents became legally enforceable as the world's first dedicated regulatory framework for AI agents.
AI agent rules: 4 jurisdictions in one week
China's agent-specific rules took effect July 15, 2026, the first anywhere, and three more jurisdictions moved within the same 7 days.
OOMeta AI Governance Platform
Vendor-independent, runtime-embedded, continuously operating AI governance. Agent Registry, compliance review, runtime governance, audit trails, covering Article 50 transparency requirements.
Book a Diagnostic