EU AI Act
Camera AI at work is in scope.
The EU AI Act can treat AI that monitors people at work as high-risk. Here is what it asks, and how Camnitive helps you answer.
The timeline
Three dates, one direction.
The Act entered into force in August 2024 and applies in stages. For camera AI, the high-risk stage matters most.
2 February 2025 · In force
Prohibited practices
Banned practices apply, including emotion recognition at work. Fines reach €35M or 7% of global turnover.
2 August 2025 · In force
General-purpose AI
Obligations for general-purpose AI models apply, and national authorities are appointed.
2 December 2027 · As amended
High-risk systems
Risk management, logging, human oversight, transparency and conformity assessment for high-risk AI, the category workplace monitoring can fall into. Moved from 2 August 2026 by the Digital Omnibus on AI.
Dates from Regulation (EU) 2024/1689 as amended by the Digital Omnibus on AI, Regulation (EU) 2026/1744, in force since 27 July 2026. High-risk AI in regulated products follows from 2 August 2028. Confirm the timetable that applies to you with your counsel.
Obligation by obligation
What the Act asks. What you’ll have.
Most of the evidence comes from normal operation: logs, reviews and explanations made as the system runs.
- Record-keeping and logging
- The Act asks: High-risk systems log events automatically, so decisions can be traced.
- With CamnitiveWith Camnitive: Each alert keeps the rule that fired, its frame and its clip. Every change and sign-in goes on a chained, tamper-evident audit log.
- Human oversight
- The Act asks: People oversee AI decisions that affect people.
- With CamnitiveWith Camnitive: People review alerts as valid, invalid, false positive or repeated, and each review is recorded. Reviews change the compliance scores.
- Transparency
- The Act asks: Outputs are clear enough for deployers to understand and act on.
- With CamnitiveWith Camnitive: Each alert names the rule and zone behind it. When the vision-language model confirms or rejects an alert, it gives its reason.
- Documentation
- The Act asks: High-risk systems are documented and assessed before use.
- With CamnitiveWith Camnitive: Rules live in FloFusion as versioned policies, so the record of what the system watches for is always current.
- Data governance
- The Act asks: Data is governed, relevant and handled with privacy in mind.
- With CamnitiveWith Camnitive: Footage is processed on site. Bystanders can be blurred in evidence copies, and access to evidence is logged.
- Prohibited practices
- The Act asks: Emotion recognition at work and untargeted biometric scraping are banned.
- With CamnitiveWith Camnitive: Camnitive watches operations, not emotions. Rules look at objects, zones and actions. They don’t identify anyone.
Beyond the EU
One design for every jurisdiction.
Footage processed on site, data kept in the country you choose and evidence sealed for auditors answer laws well beyond Brussels.
- Data protection
- Footage of staff and visitors is personal data under laws such as the GDPR, UK GDPR, Saudi PDPL, Indonesia’s PDP Law and Singapore’s PDPA.
- Residency
- Video is processed where it’s recorded. FloFusion can be hosted by Camnitive or deployed privately in your own cloud or data centre.
- Disclosure
- Sealed evidence and signed scores support audited safety disclosures, such as those under the CSRD.
Whitepaper · 14 pages
From Cameras to Cognition
How the cameras you already own detect, measure, prove compliance and show what to improve. The four steps, the economics and the security.