Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Your daily digest of the most important AI developments, Sunday, 22 March 2026. Today’s stories have something in common: AI isn’t just advising anymore, it’s acting.
The shift isn’t about smarter chatbots or better benchmarks. It’s about AI moving from answering questions to taking action, drafting and publishing web content, moderating social platforms in real time, landing in living rooms as a genuinely conversational assistant, and now performing on screen after death. The line between AI tool and AI operator is getting harder to find.
WordPress.com gives AI agents write access, on 43% of the web
WordPress.com announced on Friday that AI agents can now draft, edit, and publish content on customers’ websites, not just read it. The update extends the platform’s existing Model Context Protocol (MCP) support, which launched last autumn and allowed AI assistants like Claude Desktop and Cursor to read site content and analytics. Those same agents can now write new posts, create landing pages, manage comments, restructure categories, fix alt text, and update metadata.
That matters at scale. WordPress powers over 43% of all websites on the internet. The WordPress.com hosted platform alone handles 20 billion page views and 409 million unique visitors every month. Giving AI agents write access across that infrastructure is a real shift, websites could theoretically run with minimal human involvement beyond a natural language prompt.
There’s a human nominally in the loop: site owners set instructions via natural language commands, and changes are tracked in the activity log. But if AI agents can generate and publish content autonomously at this scale, the composition of the web changes in ways nobody has fully worked out yet.
Source: WordPress.com now lets AI agents write and publish posts, TechCrunch
Meta’s AI support assistant catches 5,000 scams a day on Facebook and Instagram
Meta rolled out its AI Support Assistant across Facebook and Instagram this week. The system handles account support from start to finish, reporting scams, tracking content takedowns, managing privacy settings, resetting passwords, processing appeals, and responds to requests in under five seconds.
The standout number: Meta says the system is now catching around 5,000 scams per day without waiting for user reports. That’s a move from reactive moderation to proactive interception. The assistant runs 24/7 in all markets where Meta AI is available, on iOS, Android, and the Help Centres on desktop.
For users it means faster resolution. For Meta it means scaling trust and safety without scaling headcount proportionally. Whether AI moderation at this volume reduces errors or compounds them at scale is a question that won’t be answered until something goes wrong.
Source: Boosting Your Support and Safety on Meta’s Apps With AI, Meta Newsroom
Alexa+ arrives in the UK as Amazon’s first European rollout
Amazon’s redesigned Alexa, rebuilt around generative AI and branded Alexa+, launched in the UK on 19 March as part of an Early Access programme. It’s the first European market to get the updated assistant, already available in the US.
Amazon says UK customers have used Alexa over 114 billion times in the past three years, making it the country’s most-used voice assistant. The new version is built to feel conversational rather than command-response: it takes action, integrates with smart home devices, and personalises over time.
Alexa+ is free during Early Access. After that: free for Prime members, £19.99 a month for everyone else. Invitations are going out to hundreds of thousands of UK users over the coming weeks.
Source: Alexa+: Amazon’s next-gen AI assistant now available in UK, About Amazon
Microsoft’s MAI-Image-2 reaches #3 globally for text-to-image
Microsoft launched MAI-Image-2 on 19 March. It immediately entered the top three on the Arena.ai text-to-image leaderboard, putting it alongside the established front-runners. The model is live on the MAI Playground now, with Copilot and Bing Image Creator rollout underway.
This generation’s focus was photorealism: accurate skin tones, natural lighting, environments that look used rather than generated. Microsoft also worked on text within images, historically a weak point for diffusion models, so MAI-Image-2 can produce infographics, slides, and posters with readable text. API access is available today for enterprise customers including WPP; broader developer access comes later.
Source: Introducing MAI-Image-2: for limitless creativity, Microsoft AI
Val Kilmer will appear in a new film, as an AI reconstruction
Val Kilmer died in 2025 after years battling throat cancer. He was originally cast as Father Fintan, a Catholic priest and Native American spiritualist, in “As Deep as the Grave,” a true story about archaeological excavations in Canyon de Chelly, Arizona. He never made it to set.
Director Coerte Voorhees has now built Kilmer’s character using generative AI, drawing on photos provided by the family from different periods of Kilmer’s life, along with audio recreations of his voice. The estate, Kilmer’s daughter Mercedes, and his son Jack all signed off on it. Voorhees says Kilmer himself had expressed a desire to be part of the film. The AI version will appear in “a significant part” of the finished movie alongside Tom Felton, Abigail Lawrie, Wes Studi, and Abigail Breslin.
The family’s approval and Kilmer’s stated wishes are doing a lot of work here ethically. The film industry hasn’t resolved what AI recreation of deceased actors should actually look like, legally, contractually, or morally. This case will come up in those conversations for a while.
Source: Val Kilmer to Star Posthumously in New Film Via AI, Variety
The pattern across all of it
AI is writing the content, moderating the platforms, answering your questions in real time, and now performing on screen after its subject has died. The jump from “AI assists” to “AI does” has been coming for a while. What’s different now is the deployment scale, 43% of the web, two of the world’s largest social platforms, millions of UK households.
The governance questions are running behind. Who is accountable when AI moderation gets it wrong at 5,000 cases a day? What happens when AI-published content at WordPress scale is indistinguishable from human-written? When does estate approval for an AI actor become a legal standard, and when does it become a loophole? These aren’t hypothetical. They’re live now.
FAQ
Can AI agents publish content on WordPress without human approval?
With the new WordPress.com update, AI agents can draft, edit, and publish posts using MCP (Model Context Protocol). Site owners provide instructions via natural language, and the agent can execute, including publishing, without step-by-step approval for each action. All changes are logged in the site’s activity feed.
Is Alexa+ free in the UK?
During the Early Access period, Alexa+ is free for invited UK users. After Early Access ends: free for Amazon Prime members, £19.99 a month for non-Prime customers.
What is MAI-Image-2 and how does it compare to other AI image generators?
MAI-Image-2 is Microsoft’s second-generation text-to-image model, launched in March 2026. It currently ranks #3 on the Arena.ai global leaderboard. Its strengths are photorealism, in-image text generation, and detailed scene composition. It’s available via the MAI Playground, Copilot, and Bing Image Creator.
Did Val Kilmer’s family approve the AI recreation?
Yes. Kilmer’s estate, his daughter Mercedes, and his son Jack all gave approval. Director Coerte Voorhees has said Kilmer himself wanted to be part of the project before his death in 2025.
That’s today’s briefing from FridayAIClub.com. We track the AI developments that actually matter, every day. Check back tomorrow for the next update, or subscribe below to get it delivered.
