No. 8Saskatoon, Saturday, October 10, 2026Price: nothing

TheThe Byte Beaver

Platform news for the people who buy the ads. Canada and the US.

8 min read

Meta bans ByteDance ads from Facebook and Instagram in 7 markets

Ad platformsData lawToolsCapital

The three that matter

  • Google Google Ads automated promotions go live this Sunday. Opt out before October 12 or Google will automatically attach website coupons and deals to eligible Search and Performance Max campaigns.
  • TikTokCanada Meta has blocked ByteDance from advertising on Facebook and Instagram in the US, Canada, and five other countries, marking the clearest competitive move Meta has made against TikTok's parent to date.
  • OpenAI GPT-6 is now available to all ChatGPT users with a new interactive Intelligent UI that generates charts, forms, and buttons inside answers, raising the bar for what users expect from AI tools.

Advertising & Ad Tech

  • Act Now Google Google Ads will auto-enroll eligible campaigns in automated promotions starting October 12. Why it matters: If you manage Search or Performance Max campaigns with location assets and no existing promotion assets, Google will start pulling offers from your website and attaching them as ad extensions without further notice. Go to Assets > Account-level automated assets > Advanced settings and set Automated Promotions to Off before Sunday. Expired deals and free-trial language can surface as public promotions if you leave it on. [Search Engine Roundtable] [Relevant Audience]

  • Heads Up TikTokCanada Meta has banned ByteDance and ads linking to TikTok or other ByteDance properties from its platforms in seven countries. Why it matters: The restriction covers the US, Canada, Egypt, Indonesia, Japan, Thailand, and Vietnam, effective immediately. Any third-party campaign that links to a ByteDance property in those markets is also affected. Meta called it standard practice for a competitor. For agencies running campaigns that touch TikTok in any of those markets, audit creative and destination URLs now. [Bloomberg] [Business Standard]

  • Heads Up MetaTools Meta launched several AI ad tools at Advertising Week New York on October 6. Why it matters: Customer Lifecycle Strategy is now open to all advertisers. Video generation from static image assets in Advantage+ Creative is generally available. Both are worth testing for clients struggling with video creative volume. Meta and Sierra also announced the Personal Agent Protocol, an open standard for AI agent purchases, with Shopify, Stripe, and Walmart as founding partners. Agentic checkout inside Meta AI is still in testing. [Relevant Audience] [Common Thread]

  • Act Now Meta US political advertiser authorization deadline is Monday, October 13. Why it matters: Meta requires US political and social issue advertisers to complete authorization before that date to run ads through the November election period. The ad restriction window runs to October 27. Miss the deadline and campaigns go dark. [Social Media Today]

  • Heads Up Meta Meta developer deadlines are approaching. Why it matters: Embedded Signup v2/v3 is deprecated on October 15. The Commerce Order Management API sunsets October 27. Meta's updated Commercial Terms take effect October 30. Teams managing Meta integrations should schedule these updates this sprint. [admakeai.com]

  • Fyi Google Google's Performance Max asset experiments are now available to all advertisers. Why it matters: The A/B testing feature for PMax creative launched broadly on October 2. Agencies running PMax campaigns can now run structured creative tests rather than relying on Google's black-box optimization alone. [Common Thread]

  • Fyi Google Google's September 2026 spam update finished rolling out on October 8. Why it matters: The update ran September 24 to October 8. If organic traffic for any client changed in that window, it is now stable enough to assess. Check Search Console for the period. [Search Engine Roundtable]

Data & Privacy Law

  • Heads Up Connecticut's 2026 privacy amendments create a categorical ban on processing minors' data for targeted advertising. Why it matters: Unlike opt-in frameworks, this prohibition applies regardless of parental consent. Advertisers with any audience targeting that could reach users under 18 in Connecticut need to review their data flows now. This is the most aggressive state-level teen ad restriction yet. [TrustArc] [Osano]

  • Heads Up Law France's CNIL is finalizing rules that would require double consent for email tracking pixels. Why it matters: If adopted, open-rate tracking in email campaigns sent to French contacts would require a second affirmative consent step beyond the current single opt-in. Email marketers serving European lists should watch for the final guidance. [Secure Privacy]

  • Fyi Law The EU AI Act became generally applicable on August 2, 2026. Why it matters: High-risk system obligations are pushed to December 2027, but the framework is now in force. Agencies building or deploying AI for ad decisioning should flag this for legal review. [Pearl Cohen]

  • Fyi Law New York requires conspicuous disclosure when ads use synthetic performers. Why it matters: Senate Bill S8420A applies to AI-generated spokespeople and actors in ad creative. Agencies producing AI video ads for New York audiences should add disclosure language now. [O'Melveny]

WordPress

  • Act Now OpenAIWordPress WordPress 7.1.3 was released October 6 with seven security fixes. Why it matters: The release patches a stored XSS on the Comments admin page (exploitable via pending comments, found by Trail of Bits with OpenAI), a second-order SQL injection in WXR export, and a denial-of-service issue in the HTTP utility method. Unauthenticated disclosure of comments on private posts is also fixed. Update all WordPress installs immediately. Patchstack notes branches 4.7 through 6.5 were not yet patched as of its publication, so check older installs. Fixes have been backported to most supported branches. [WordPress.org] [Patchstack] [Pantheon]

OpenAI / ChatGPT

  • Fyi OpenAI OpenAI launched GPT-6 for all ChatGPT users on October 7, along with Intelligent UI. Why it matters: Intelligent UI generates interactive elements inside answers, including charts, forms, and buttons. The Sol model is live on paid tiers; the Luna model is rolling out to Free and Go users. This raises user expectations for AI-generated outputs and is relevant context for anyone building AI-assisted ad copy or reporting tools. [OpenAI] [9to5Mac]

  • Fyi OpenAI OpenAI banned Russian and Iranian networks that used ChatGPT to plant fake news in real media outlets. Why it matters: The Iranian operation ran seven fake journalist personas that pitched articles to small and medium outlets worldwide. OpenAI says roughly 100 AI-generated articles ran across at least 20 news outlets. This is relevant for agencies advising on brand safety and media placements: AI-authored content is now appearing in editorial environments without clear attribution. OpenAI has now disrupted 30 influence operations over two and a half years. [NPR] [Washington Post] [Euronews]

Anthropic / Claude

  • Fyi Anthropic Anthropic released Claude Haiku 5.5 on October 7, priced at roughly 75% less than Haiku 4.5. Why it matters: The price drop continues the AI cost compression trend. For early-stage builders or agencies prototyping AI ad tools, Haiku 5.5 makes high-volume inference significantly more affordable. [Yahoo Finance]

  • Fyi Anthropic Anthropic updated its usage policy on October 8 to explicitly ban election interference, weapons software, and prolonged verbal abuse of Claude. Why it matters: The election interference restriction is specifically relevant ahead of November. It applies to any business or developer using the Claude API to build tools with electoral content. [TechCrunch]

  • Fyi Anthropic Anthropic expanded access to its top-tier models for verified cybersecurity organizations. Why it matters: Opus 5.5, Sonnet 5.5, and Mythos 5.1 are now available to qualifying cyber firms alongside the US government. This is background context for the enterprise AI security conversation. [Bloomberg]

Supabase

  • Fyi Supabase Supabase raised $150 million and announced the acquisition of Turso on October 2. Why it matters: The funding round was led by GIC with CapitalG, IronArc, and SquarePeg. Supabase said the money accelerates agent-driven database development and provides employee liquidity. The Turso acquisition brings a distributed SQLite product into the Supabase ecosystem. Relevant context for early-stage builders choosing a backend. [The SaaS News] [Supabase Blog]

  • Act Now Supabase Supabase is removing auto-grants for new public schema tables on October 30. Why it matters: After October 30, newly created tables in existing projects will not be automatically exposed to the Supabase Data API. Any migration or deployment that creates tables will need explicit grants added beforehand or the tables will be inaccessible to the API. Review your migration scripts before the deadline. [DEV Community] [Supabase Changelog]

Big Tech

  • Fyi MicrosoftApple Microsoft unveiled the Surface Laptop Ultra running Nvidia's RTX Spark N1X chip, alongside an on-device AI coding model. Why it matters: Microsoft is positioning high-end Windows laptops as AI workstations capable of running coding agents locally. The announcement also included security tooling to prevent AI agents from accessing data outside their scope. This is background context for the on-device AI trend that Apple is also pursuing with its updated Mac lineup. [Reuters/Investing.com]

Sources

Every item above links to where it came from. Nothing runs that was not read first. How sourcing works. Found an error? Tell us and it gets corrected here, in the open.