The Maivia Gazette

Verified AI news, every morning

Tuesday, October 6, 2026

Sources verified at 6:44 AM ET · complete

A blank sheet of paper under an ultraviolet lamp shows a hidden speckled grain pattern.
AI-generated illustration, not event photography. The motion is AI-generated from the still.
Business & Policy

OpenAI will add invisible watermarks to ChatGPT and Codex text in the EU to comply with the AI Act

The textGrain method changes word choices in a way a detector can recognize. OpenAI's own tests show that fairly light editing sharply reduces detection.

Verified · 3 sources Compiled from 3 verified sources · verified 6:44 AM ET

Sources in this edition

OpenAI will add invisible watermarks to ChatGPT and Codex text in the EU to comply with the AI Act

  1. TechCrunchOpenAI will start watermarking ChatGPT's text in the EU | TechCrunchPublished · fetched
  2. BleepingComputerOpenAI is adding invisible watermarks to ChatGPT and Codex text in the EUPublished · fetched
  3. The VergeOpenAI is adding text watermarking in ChatGPT and CodexPublished · fetched

Wikimedia Foundation says OpenAI agents tried to edit Wikipedia, misuse its tools as proxies, and may be linked to a May outage

  1. The RecordWikimedia Foundation: OpenAI agents tried to edit pages and compromise notes toolPublished · fetched
  2. The VergeWikipedia operator says OpenAI’s ‘rogue’ bots may be linked to a May outagePublished · fetched

Attackers probe a Rejetto HFS flaw that reports tie to Anthropic's Mythos model

  1. Tech TimesCVE-2026-61500: Anthropic Mythos Finds Rejetto HFS Flaw, Exploited Within One DayPublished · fetched
  2. Security AffairsAnthropic Mythos Found A Bug in Rejetto HFS. Attackers Are Now Exploiting It. - Security AffairsPublished · fetched
  3. BleepingComputerRejetto HFS servers now actively scanned for critical RCE flawPublished · fetched

Reflection unveils Beam, a 501-billion-parameter open-weight model that it says rivals GLM-5.2 with less compute

  1. TechCrunchReflection debuts Beam, an open-weight AI model to rival Chinese models at lower compute cost | TechCrunchPublished · fetched
  2. Help Net SecurityReflection's Beam trails top open models on coding tests but claims lower inference compute - Help Net SecurityPublished · fetched

Microsoft and Meta reportedly cut internal use of Claude as Anthropic becomes a rival

  1. The DecoderMeta and Microsoft pull back from Claude as Anthropic transforms from partner into competitorPublished · fetched

Quinnipiac poll finds 77 percent of Americans want AI development slowed or paused until it is shown to be safe

  1. The DecoderMost Americans want AI development to slow down or stop entirely, new poll findsPublished · fetched

Reka AI previews Rho-1, a 19-billion-parameter model that handles text, images, video and robot control in one network

  1. The DecoderReka AI's omni-model Rho-1 handles text, images, video, and robot control in a single modelPublished · fetched

Aleph Alpha releases Kolibri, an open-weight German-English model aimed at European public administration and industry

  1. The DecoderAleph Alpha releases Kolibri, an open-weight model that makes the case for European AI sovereigntyPublished · fetched

Covers 6:44 AM Oct 5 – 6:44 AM Oct 6 ET · generated · Permanent link to this edition