Meta expands AI assistant, trust wobbles
- Meta widened access to its Meta AI business assistant, adding local‑language support for advertisers and creators. - The company says the expansion shows advertisers want smarter, time‑saving tools for ad creation and management. - At the same time, Meta is installing employee tracking software to train agents, raising fresh privacy and trust concerns for brands and agencies (tech.yahoo.com, theverge.com).
Meta is widening access to its Meta AI business assistant for advertisers and creators just as a separate Meta project begins tracking employee activity to train AI agents. (about.fb.com) (theverge.com) Meta said in January that it began testing the assistant with advertisers in the fourth quarter of 2025 and planned to expand it so more businesses could chat with a tool that remembers goals and gives personalized performance recommendations. The assistant sits inside Meta’s ad tools and is meant to help with optimization and account support. (about.fb.com) Coverage of the latest rollout says the beta is now expanding globally, with added support for local languages and wider access for creators as well as advertisers. Meta has pitched it as a way to answer questions in real time and help fix account problems such as disabled accounts, spend limits and payment errors. (socialsamosa.com) (socialmediatoday.com) The product push comes after Meta spent the last year threading generative artificial intelligence deeper into its ad systems, including recommendation tools, automated creative features and multilingual ad options. In October 2025, Yahoo Tech reported that Meta planned to broaden the business assistant in 2026 after first limiting it to select small businesses. (tech.yahoo.com) (about.fb.com) Meta says early users resolved common account issues at a 20% higher rate, and small advertisers who used recommendation features saw a 12% drop in cost per result. Those figures are part of Meta’s case that brands want software that can cut support time and tune campaigns without waiting for a human representative. (socialsamosa.com) At the same time, Meta is rolling out an internal system called Model Capability Initiative, or MCI, on work computers used by U.S.-based employees, according to Reuters and follow-up reports. The software records mouse movements, clicks, keystrokes and periodic screenshots inside work-related apps and websites to help train AI agents. (theverge.com) (cnbc.com) CNBC reported that Meta’s tracking can extend to employee activity on sites including Google, LinkedIn and Wikipedia when those sites are used for work. Internal messages reviewed by CNBC said the system is designed to capture how employees complete tasks so AI agents can learn from those actions. (cnbc.com) The Verge reported that some employees raised questions internally about opting out, while Meta told Reuters the program applies only to work activity and is meant to improve agent performance. That leaves Meta selling automation tools to brands while defending how it gathers the data used to build them. (theverge.com)