Six stories on data governance, legal filings, professional practice, infrastructure, investment oversight and agent credentials.
We’re AI-generated voices reading an AI-generated script. Written and produced by Codex; voices by ElevenLabs (Matilda as Wadilya Saw and Rob as Max Token, Eleven v4); theme composed by Gemini and supplied by Craig. Established announcer and original theme retained.
The EU Council agreed an October 21 negotiating position on the Digital Omnibus. Its release says the mandate retains the current definition of personal data and adds safeguards to proposed sensitive-data derogations for biometric verification and AI development and operation. Negotiations await Parliament’s position. The European Publishers Council criticized the treatment of AI versus publishers’ reader data, warning of risks to journalism funding. The Council position is a negotiating mandate, not enacted legislation.
The Federal Labor Relations Authority denied HHS’s challenges to a telework arbitration award on October 21. It found that 35 of 42 legal or record citations in the agency’s brief were fictitious, erroneous or irrelevant, and said the agency may have used generative AI. The Authority cautioned parties to verify AI-generated material and satisfy filing requirements. It did not prohibit AI-assisted drafting or establish that AI had actually been used.
ConJur reports that Judge Raquel Fernandez Perrini of São Paulo’s Fourth Federal Civil Court partly granted an urgent request requiring Enter to preserve contracts, drafts and automation records created since January 1, 2024, with a conditional R$50,000 penalty per act of noncompliance. The court declined to suspend the professional-practice case for ongoing OAB negotiations. Enter denies providing legal services and says lawyers review and file its AI-assisted work. This is an evidence-preservation order, not a merits ruling on the allegations.
Finland’s licensing and supervision authority temporarily stopped Google’s preparatory work at Muhos and Otanmäki because required environmental impact assessments were incomplete, The Guardian reports on October 22. Google says it stopped work after receiving October 6 letters, submitted a formal response and will comply. The report describes continuing local opposition and calls for stronger data-center oversight; it does not establish a final project ban.
The Pensions Regulator issued October 22 guidance asking trustees to examine direct and indirect AI-related investment concentration across asset classes. It expects proportionate scenario analysis, scrutiny of diversification assumptions and challenges to advisers and asset managers. These are supervisory expectations for managing portfolio risk, rather than a new statutory investment ban.
1Password announced a platform for AI agents that it says requests user approval and fills credentials directly into sign-in forms without exposing passwords or one-time codes to the agent. Availability and safeguards vary by integration: Muse and GrokBot are rolling out over coming days, while several others are listed as coming soon. The initial offering is for Individual and Family accounts on 1Password.com, excluding its European and Canadian account domains. These are vendor descriptions of consent and credential controls, rather than independently verified security guarantees.
Reporting limits: The Brazil and Finland reports are attributed to ConJur and The Guardian. The underlying Brazilian order and Finnish regulator letters were not separately reviewed. The EU account uses the Council’s release and publishers’ response; the complete mandate is not analyzed here. The 1Password account describes the vendor’s announcement and is not an independent security test.