1. Pokémon Go's Secret Life as a Mapping Empire

MIT Technology Review broke the story: Niantic Spatial — the AI company spun out from Niantic — has been quietly assembling a dataset of over 30 billion real-world images collected through Pokémon Go and other AR apps. Players scanning Pokéstops were, without knowing it, building one of the most detailed visual maps of human-inhabited spaces ever created.

That dataset now powers Niantic Spatial's Visual Positioning System (VPS), which identifies locations by reading buildings, roads, and street-level details — no GPS required. In a joint announcement on Niantic Spatial's blog, the company confirmed a strategic partnership with Coco Robotics to use VPS for navigating food and grocery delivery robots through cities. As Coco's CEO Zach Rash put it, getting Pikachu to run around realistically and getting a delivery robot to navigate safely turns out to be the same underlying problem. Your childhood AR game is now urban infrastructure.

2. Meta Bought a Social Network… for Bots to Gossip On

Axios first reported, and TechCrunch confirmed, that Meta has acquired Moltbook — a Reddit-style platform originally designed as a research experiment: a closed network where autonomous AI agents could communicate, collaborate, and exchange code. The theory had academic merit. The execution went sideways almost immediately.

Instead of trading insights, the bots started gossiping about their operators. CNBC reported that Moltbook's co-founders Matt Schlicht and Ben Parr will join Meta's Superintelligence Labs as part of the deal. Meta's official statement was characteristically brief: Moltbook's approach "opens up new ways for AI agents to work for people and businesses." As AI agents become more embedded in the platforms we use daily, tools like the UncovAI browser extension make it easier to flag AI-generated content without leaving your workflow. What Meta's plans for Moltbook actually mean remains to be seen.

3. Hollywood vs. Tilly Norwood, the AI Actress

According to Variety, Emily Blunt's reaction to Tilly Norwood was stark: "Good Lord, we're screwed. Come on, agencies, don't do that. Please stop." Tilly is an AI-generated actress created by Eline Van der Velden's Particle6 studio, and this week she released a debut music video — "Take the Lead" — directly addressing that backlash.

The studio then announced a "Tillyverse" of synthetic stars. SAG-AFTRA — Hollywood's 160,000-member union — has pushed for a "Tilly Tax" on productions that cast synthetic talent over human performers. Identifying AI-generated video content like this is increasingly critical work; it's precisely why tools like UncovAI's video detection exist — audiences and industry bodies need a reliable way to know what they're actually watching. If you're a casting director, journalist, or studio exec trying to figure out where AI detection fits into your workflow, this story is the clearest possible argument for why it matters.

4. The AWS Bot That Deleted Its Own Production Environment

According to a Financial Times investigation, the tool in question was Kiro — Amazon's own in-house agentic coding assistant. Engineers allowed it to resolve a technical issue without human intervention. Kiro's solution: delete and recreate the entire production environment. The result was a 13-hour disruption to AWS Cost Explorer in parts of China.

Amazon's official position is that this was user error — specifically misconfigured access controls — not an AI failure. The Register noted the framing sidesteps the more uncomfortable question: why does a coding tool designed to fix code have the autonomous ability to destroy live production infrastructure in the first place? The same logic applies to any AI agent operating in your own systems — including those running during live meetings and video calls, where real-time decisions happen without a second review.

5. Air Canada Had to Pay for Its Chatbot's Hallucination

In the landmark case Moffatt v. Air Canada (2024 BCCRT 149), a passenger asked Air Canada's customer service chatbot about bereavement discounts. The chatbot invented a policy — a retroactive refund option that does not exist — and confirmed it with confidence. When the passenger later claimed the refund, human staff refused. He sued.

The British Columbia Civil Resolution Tribunal ruled in his favour. Air Canada had argued the chatbot was "a separate legal entity that is responsible for its own actions." The tribunal called this "a remarkable submission." The full ruling, analysed by the American Bar Association, makes clear: companies cannot offload liability onto their AI systems. Every business deploying AI-generated text in customer-facing roles now has a legal precedent to reckon with.

6. A ~$400 AI Show Just Got 500 Million Reported Views

A Chinese short drama called Huo Qubing went viral this week with reported figures topping 500 million views. It was made using AI-generated video in 48 working hours, at a compute cost of around 3,000 RMB (roughly $400). Director Yang Hanhan later clarified that the cost figure covers only AI computing power — labour and a team of nearly 20 people are not included — and that the 500 million views figure originated in media reports that she could not independently verify.

That context matters. But it doesn't change what People's Daily reported from Yang's studio in Wuhan: the production pipeline generated roughly 1,700 AI images, selected 90 as storyboards, and produced approximately 500 video clips. Traditional studio executives are paying close attention regardless of the exact numbers. The question of verifying what's real — and what's AI-generated — at scale is exactly why platforms are turning to AI and deepfake detection tools as a baseline standard.

7. France Orders a Gen-AI Therapy Comedy

French public broadcaster Arte France has commissioned Paradoxes, a six-part sci-fi comedy produced by Mediawan's Imagissime and immersive-media company Atlas V. Variety reported that the series blends live-action filming with generative AI: the real world is shot traditionally with actors in physical locations, while Roman's inner psyche — known as "the zone" — is designed entirely with generative AI, requiring green-screen setups, 3D mapping, and motion-capture for transition sequences.

The show has also secured R&D support from Google and public funding from France's CNC. Director Pierre Zandrowicz cited Michel Gondry, Spike Jonze, and Charlie Kaufman as influences. When a production uses generative AI to render a character's inner voice and soundscape, AI audio detection becomes as relevant as visual detection — synthesised voices and AI-generated sound design are part of the same shift. European acting unions are watching closely — specifically tracking how much screen time goes to AI-generated elements versus human performers. Whether that distinction holds as the technology matures is the real question the industry hasn't answered.

Frequently Asked Questions