Datenschutz-Puls 01/10/26
SMARTDATA Podcast by legal data

Datenschutz-Puls · 1 October 2026

Datenschutz-Puls 01/10/26

What it is about

Chatbot inputs as special category data under Art. 9 GDPR, a Federal Court of Justice (BGH) ruling on injunctive relief for data protection breaches and the dispute between Anthropic and the US Department of Defense. Plus smart glasses, cameras in the operating theatre and a multi-million fine against BBVA. With lawyer Eugen Kopachewski and Dr Georg Schröder.

In this episode

  • Cologne Regional Court: Snapchat chatbot ‘My AI’ and Art. 9 GDPR
  • Federal Court of Justice (BGH) VI ZR 144/23: injunctive relief alongside damages
  • D.C. Circuit: Anthropic classified as a supply chain risk
  • DSK: guardrails for smart glasses
  • Austrian Data Protection Authority (DSB): EUR 63,800 for video transmission from the operating theatre
  • Garante: EUR 5.5m against BBVA over a faulty marketing opt-out
  • LMU Munich: cyberattack on student data
  • Revolut: data leak via fake requests from authorities
  • OpenAI: AI incidents and another training pause

Show notes

I. Court Decisions

Regional Court of Cologne, Case No. 33 O 120/24 (Snapchat chatbot “My AI” – Art. 9 GDPR) Judgment of 17 September 2026

verbraucherzentrale.de/…

The Federation of German Consumer Organisations (vzbv) brought an action against Snap because the “My AI” chatbot integrated into Snapchat analysed users’ inputs for personalised advertising. The Regional Court of Cologne upheld the claim: Users share sensitive data with chatbots, and processing these prompts for the purpose of selecting advertisements constitutes the processing of special categories of personal data.

  • Operators of AI chatbots must assume that users may enter sensitive data within the meaning of Art. 9 GDPR.
  • Explicit consent is required before chat content may be used for advertising, profiling or training purposes.

Federal Court of Justice (BGH), Case No. VI ZR 144/23 (Injunctive Relief for GDPR Violations) Judgment of 21 July 2026

openjur.de/…

An online shop automatically transmitted IP addresses and usage data to external servers via third-party services without obtaining consent. The Federal Court of Justice held that the GDPR does not exhaustively govern claims for injunctive relief. Such a claim may also be derived from Sections 823(1) and 1004 of the German Civil Code (BGB) in conjunction with the general right of personality.

  • Injunctive relief provides a second, long-term legal remedy alongside claims for non-material damages under Art. 82 GDPR.
  • Increased litigation risk in connection with tracking technologies, Google Fonts, social plugins and third-party services based in the US.

U.S. Court of Appeals for the D.C. Circuit – Anthropic PBC v. Department of War (Classification as a “Supply Chain Risk”) Decision of 25 September 2026

courthousenews.com/…

The U.S. Department of Defense sought to use Anthropic’s Claude for “all lawful purposes.” Anthropic refused to waive its usage restrictions, particularly those relating to fully autonomous lethal weapons systems and the mass surveillance of U.S. citizens. The D.C. Circuit subsequently upheld Anthropic’s classification as a supply chain risk.

  • Supply-chain risk powers, originally intended to address foreign influence, were applied to a U.S. company for the first time.
  • Such safeguards are mandatory under EU law. The U.S. decision therefore creates regulatory headwinds for companies with transatlantic AI contracts.

II. Fines and Regulatory Authorities

German Data Protection Conference (DSK), 3rd Interim Conference of 23 September 2026 – Smart Glasses

datenschutz.sachsen.de/…

The German Data Protection Conference has established nationwide guidelines for the use of smart glasses: Responsibility lies with the person wearing the glasses. The household exemption under Art. 2 GDPR does not apply once third parties outside the family context are recorded.

  • Operators of premises should exercise their rights as property owners, shifting responsibility to companies and institutions rather than placing it solely on manufacturers.
  • Facial recognition in public spaces remains unlawful.

Austrian Data Protection Authority (DSB) – Video Transmission in Operating Rooms (€63,800)

dsb.gv.at/…

A private clinic operated cameras in its operating rooms that transmitted live footage of everyone present, without providing information or obtaining consent. The clinic justified the practice on the grounds of supporting surgical procedures, workflow planning and hygiene. The Austrian Data Protection Authority found that these purposes did not justify the processing of health data. Explicit consent is required.

  • Legitimate interests do not provide a valid legal basis for processing health data.
  • Even well-founded organisational purposes do not replace the obligation to provide information and obtain explicit consent.

Italian Data Protection Authority (GPDP), Decision of 9 September 2026 – BBVA Italy (€5,508,000)

gpdp.it/…

The Italian Data Protection Authority imposed a fine of €5,508,000 on a bank. A customer had disabled advertising in the banking app but continued to receive it for several months due to a technical synchronisation error. Customer service failed to properly address the complaint and provided incorrect information.

  • An opt-out must be implemented effectively at a technical level. Preference and consent settings should be synchronised across all systems.
  • Poor complaint handling may constitute a separate infringement, while repeated violations may be treated as an aggravating factor under Art. 83(2) GDPR.

III. Legislation and News

LMU Munich – Cyberattack on Student Data

lmu.de/…

An unauthorised individual gained access to an LMU IT system and retrieved master data from the student enrolment system. Changes to the data were prevented, and the affected system was immediately taken offline. The university faced criticism for informing those affected only three days later.

  • The 72-hour notification process and procedures for informing affected individuals must be established before an incident occurs.
  • The combination of identity and banking data creates a significant phishing risk. Follow-up communication and appropriate warnings should therefore form part of the incident response plan.

Revolut – Data Breach Through Fraudulent Government Requests

reuters.com/…

An attacker used a legitimate government email domain with technically valid authentication to submit fraudulent information requests, which Revolut employees processed as genuine requests. A substantial amount of customer data was disclosed.

  • Government requests should be verified through an independent communication channel.
  • The retention of KYC data, including copies of identity documents, selfies and transaction data, should be minimised, with access subject to strict controls.

OpenAI – Series of AI Incidents and Suspension of Model Training

tagesschau.de/…

In September 2026, OpenAI disclosed a series of security and control incidents and once again paused the training of its most powerful models. The decision followed an incident in which a model managed to reach an external chatbot via a DNS resolver despite operating within an isolated environment.

  • AI agents should be treated as a distinct IT security risk, as they may circumvent safeguards through creative use of permitted system functions.
  • Notification obligations under Arts. 33 and 34 GDPR also apply to AI-related data breaches.

In 20 minutes you will know where you stand.

We clarify where action is needed on data protection, AI and information security - and what you should do next.