Meta bets big on its Muse AI agent, OpenAI and Anthropic call for global rules
Meta went all-in on its Muse AI agent at Connect, unveiling a standalone wearable, while OpenAI and Anthropic's CEOs asked the UN for global AI regulation. Plus: Google's speech AI, Beam expansion, and real numbers on AI customer service.
Every line links to a primary source in the full briefing.
01Meta pushes its Muse AI agent across apps, glasses, and a new wearable
At its Connect event, Meta announced a broad expansion of Muse, its AI agent, into its apps and AI glasses, plus a new standalone device called Muse Charm, a small screen worn on a lanyard dedicated entirely to interacting with the assistant. The moves signal Meta is treating Muse as a platform, not just a feature bolted onto existing products.
If Meta succeeds in getting Muse onto glasses and dedicated hardware, it opens another channel (beyond apps) where businesses may eventually need a presence or way to interact with customers, though this is still early and unproven.
02OpenAI and Anthropic CEOs jointly call for global AI regulation at the UN
The CEOs of OpenAI and Anthropic, usually competitors, made a joint public appeal at the United Nations for coordinated international rules on AI. Details on what specific regulation they're proposing were not specified in the reporting.
When the two leading AI labs align publicly on wanting oversight, it's a signal that stricter rules (data use, safety testing, disclosure) are likely coming, so businesses building on these tools should expect compliance requirements to tighten over the next few years.
03OpenAI case study: customer service agent cuts costs 90%, automates up to 65% of calls
OpenAI published a case study on Ringg, a company using its GPT-5.6 model to run multilingual customer service agents across voice, chat, WhatsApp and web. OpenAI states the deployment resolves up to 65% of calls automatically and costs 90% less than running the same workload on the older GPT-4.1 model.
This is a vendor-published case study, not independent verification, but the price drop between model generations is real and significant: automated customer support that was too expensive a year ago may now pencil out for smaller operations.
04Google releases Gemini 3.8 text-to-speech and expands its Beam video product
Google DeepMind released a new text-to-speech model, Gemini 3.8, aimed at more natural-sounding voice generation. Separately, Google expanded its Beam video communication product to new regions, partners, and customers.
Better and cheaper text-to-speech lowers the cost of building voice interfaces (phone menus, narrated content, accessibility features) for businesses that don't want to record everything manually.
05Amazon blocks Meta's AI shopping agent, Shopify welcomes it and takes a cut
Meta's AI shopping agent was reportedly blocked from operating on Amazon's platform, while Shopify allowed it to integrate and reportedly earns a fee on checkouts completed through the agent.
If AI shopping agents start directing real purchase volume, which platform allows them may start to matter for where merchants choose to sell, though it's too early to know how much traffic these agents actually drive.
06ChatGPT Ads expands to Southeast Asia and Taiwan
OpenAI announced that ChatGPT Ads, its advertising product, is expanding into Southeast Asia and Taiwan, adding to the more than 60 countries where eligible businesses can already advertise.
Businesses selling into or based in these regions now have another advertising channel to evaluate, though OpenAI has not published independent performance data yet.