🧔♂️ A friendly human may check it before it goes live. More news here
Italy closes probe into DeepSeek after AI ‘hallucination’ warnings
Italy’s antitrust regulator has ended its investigation into DeepSeek, an AI system operated by Chinese firms Hangzhou DeepSeek Artificial Intelligence and Beijing DeepSeek Artificial Intelligence, after the companies agreed to improve warnings about the risk of AI-generated false information.
The probe began in June 2025 over concerns that DeepSeek did not sufficiently inform users that its AI could produce inaccurate or misleading content, sometimes called “hallucinations.”
The Italian authority said in its latest bulletin that DeepSeek will now make these disclosures clearer and more immediate for users.
DeepSeek’s commitments include a set of measures designed to better inform people about the risk of receiving fabricated or erroneous outputs when using the AI tool.
The case was closed following the acceptance of these binding commitments.
🔗 Source: Reuters
🧠 Food for thought
Implications, context, and why it matters.
Italy sets a precedent for EU-wide AI transparency rules
- Obligations for general-purpose AI (GPAI) model providers began on August 2, 2025 1, while Article 50 marking and labeling duties start in August 2026 2. Providers must keep technical documentation current, then share it with downstream deployers (companies that integrate the models into their products) or regulators on request 3.
- The EU’s GPAI Code of Practice from July 2025 is voluntary 3. It asks providers to share technical documentation, including intended uses, with downstream users like developers integrating the model; it also requires information for regulators on request and safeguards against infringing content.
- The European Commission is drafting a Code of Practice on marking and labeling AI-generated content 2. Working groups are defining how providers and deployers (organizations that implement AI systems in applications) should disclose AI-generated content to natural persons (individuals).
AI compliance tools can serve fragmented disclosure rules
- Software vendors can build jurisdiction-specific disclaimer systems that adjust warnings by user location 3, while the EU GPAI Code of Practice asks providers to keep documentation current and share updates with deployers, which drives demand for automated tracking 1.
- A/B testing and user experience firms can tune disclosure placement to balance compliance with usability as working groups refine requirements before August 2026 2.
- Consultancies can bundle advice on the GPAI Code of Practice’s copyright and safety duties 3. The duties cover external audits and post-market monitoring for systemic-risk models (systems whose scale and capabilities may pose broader societal or market risks).
