AI journal · every morning

Today in AI

AI text in parliamentary letters, Claude Code leaks passwords via settings file

Ministers sent the Dutch House of Representatives 76 documents containing AI-generated text, without disclosing it

De Volkskrant examined 3,597 letters to the House of Representatives and answers to parliamentary questions from the Jetten cabinet, sent between 23 February and 24 October. In 76 documents, text generated wholly or partly by AI was present, two letters consisted entirely of it, and in not a single case was it stated that AI had been used. In some cases civil servants used the free version of ChatGPT, which runs counter to the Dutch central government's own rules.

What does this mean? If you work for a government body or for a client with AI rules: use only the approved business environment, keep a record of how you used AI, and disclose it where that is expected.

Government & municipalities de Volkskrant (via ANP)

Claude Code stores approved commands containing passwords, 1,122 files publicly exposed on GitHub

If you choose "Yes, and don't ask again" in Claude Code, the tool saves the approved command word for word in .claude/settings.local.json, including any passwords or API keys it contains. Security firm Apiiro found real secrets in 1,122 of 104,185 public copies of that file on GitHub, among them 379 database passwords and 23 AWS keys. Deleting the file later does not help, because older versions remain in the git history.

What does this mean? If you use Claude Code: add .claude/settings.local.json to your .gitignore, use environment variables instead of passwords in commands, and replace any keys that have already ended up in a repository.

IT & software development Cybersecurity The Hacker News

A civil servant cannot simply refuse AI, but need not cooperate with unlawful use

Binnenlands Bestuur set out what a civil servant who is opposed to AI in principle may do. An employer may use its right to give instructions to decide which tools you work with, so you cannot simply refuse an AI system approved by the municipality on principle. Refusal is possible, however, if a task conflicts with laws and regulations such as the GDPR, if there are concrete risks of errors, discrimination or privacy violations, or if the employer fails to provide sufficient training, which the EU AI Act requires.

What does this mean? If AI is made mandatory at your job, ask about the agreements on privacy and training, and report it when you see a system making mistakes.

Government & municipalities HR & personnel Binnenlands Bestuur

Every Friday: what changed in AI for your field

New models, prices, oversight and tools, with sources and what it means for you. One email a week, unsubscribe with one click.

View a sample

Your fieldEmail address

We only use your address for this newsletter. You confirm first by email.

Previously published

All editions of the AI journal and all articles, newest first. 150 posts.

9 November 2026 · ArticleAI text in parliamentary letters, Claude Code leaks passwords and can you refuse AI?Three posts from recent days with the same lesson: AI at work calls for agreements. De Volkskrant found AI text in letters from ministers, an instruction file from Claude…8 November 2026 · AI journal · 3 postsFlaw in ChatGPT Lockdown Mode patched, Suno back in courtA patched leak in ChatGPT's security mode, new lawsuits against music generator Suno, and Zwolle deploying AI agents for civil servants.8 November 2026 · ArticleAI agents hunt leaked keys, Gemini video 30% cheaper and Perplexity launches TeammatesThree posts from this week that you will notice in your work. Microsoft sees attackers using an AI agent to exploit keys leaked on GitHub, Google lowers the price of video generation …7 November 2026 · AI journal · 5 postsCabinet presents AI vision, NCTV: AI makes cyberattacks faster and largerThe cabinet charts its own course on AI, the NCTV warns that AI is speeding up cyberattacks, and Microsoft limits the free Copilot Chat at work.7 November 2026 · ArticleCabinet sets AI course, NCTV warns of faster attacks and Copilot Chat gets limitsThree posts from the past two days that bear directly on how you work with AI at work. The cabinet is setting its own AI course, the NCTV sees attackers using AI at every stage …6 November 2026 · AI journal · 3 postsFabricated sources in Accenture reports, Gemini Omni 1.2 Flash makes longer videosGPTZero finds fabricated sources in three Accenture publications, Google releases a cheaper video version of Gemini, and OpenAI does not go ahead with the price cut for ChatGPT Pro 200.6 November 2026 · ArticleClaude links bank accounts, GPT-6 Cyber for security professionals and Meta buys ConductorThree stories from the past two days. Anthropic lets Claude look into your bank account, OpenAI releases a separate model for security work, and Meta takes over a tool …5 November 2026 · AI journal · 3 postsClaude watermark now also on older models, ChatGPT makes GIFs with Images 2.6Since 3 November, Anthropic has also been marking text from Mythos 5, Opus 4.7 and Sonnet 4.6, OpenAI is launching Images 2.6, and BREIN is opening a reporting point for deepfakes.5 November 2026 · ArticleClaude watermark on older models, message threads without ZDR and more Codex in ChatGPT ProFour posts from the past few days. Anthropic now also marks text from older Claude models, the new message threads in the Claude API fall outside Zero Data Retention, Ope…4 November 2026 · AI journal · 3 postsGrok text gets a watermark for the AI Act, EU adviser critical of GeminiSpaceXSI marks Grok output invisibly from 7 November, an adviser to the EU Court sees possible infringement by Gemini of press rights, and the UWV sees more higher-educated people on unemployment benefits due to AI.4 November 2026 · ArticleGrok gets a watermark, EU adviser critical of Gemini and more highly educated people on unemployment benefitsThree updates from the past 48 hours that you'll notice if you work with AI professionally. Grok text from the API becomes recognisable as AI-generated from 7 November, an adviser to the EU...4 November 2026 · ArticleChoosing AI for cybersecurity: the complete guide (November 2026)A security team receives thousands of alerts every day from endpoints, network, cloud and email. AI helps you sift through that stream, speed up investigations and write reports.…