The announcements spanned consumer platforms, workplace software and governance frameworks, showing how AI is being layered into day-to-day experiences across industries. The focus was not only on adding new tools, but also on embedding AI more deeply into existing ecosystems, shaping how users create, work and interact online. With each update, the world’s largest firms are signaling that AI is now central to their growth and operational priorities.
Meta Bets on AI-Generated Video
Meta introduced Vibes, an AI-powered short-form video feed inside the Meta AI app and on meta.ai, according to a Thursday (Sept. 25) press release.
Users can generate clips from prompts, remix existing videos by altering visuals or music, or rework their own content. The resulting videos can be posted to Vibes or cross-shared to Instagram and Facebook Stories or Reels, per the release.
“The feed will become more personalized over time…,” the release said.
“We’re working on even more powerful creation tools and models with a number of talented visual artists and creators and will be rolling these out more widely in the future,” the release added.
The rollout also reflects broader changes across Meta’s platforms. Instagram has been shifting its strategy to emphasize video as it competes more directly with TikTok. That context makes Vibes not just an experiment in generative content but part of Meta’s ongoing effort to keep users engaged in an increasingly video-first social ecosystem.
Microsoft Moves From Office Assistant to Agent
Microsoft rolled out vibe working, expanding its Copilot suite with Agent Mode for Excel and Word and Office Agent in Copilot chat, according to a Monday (Sept. 29) company blog post.
Agent Mode in Excel can “generate outputs [and] evaluate results, fix issues and repeat the process until the outcome is verified,” the post said.
Meanwhile, “Agent Mode in Word turns document creation into vibe writing, an interactive, conversational experience,” per the post.
Office Agent allows users to request complete Word documents or PowerPoint decks from a prompt, with the AI clarifying intent and producing polished outputs, according to the post.
The launch marks a step toward embedding more autonomous AI into everyday office tools, positioning Copilot as a collaborator rather than just a helper, Axios reported Monday.
At the same time, Microsoft is rebalancing its AI partnerships. The company has been deepening its ties with Anthropic as part of a broader move to diversify away from sole reliance on OpenAI.
Google DeepMind Sharpens Safety Lens
DeepMind released version 3 of its Frontier Safety Framework Sept. 22, adding a new category for harmful manipulation, defined as the ability of models to shape beliefs or behaviors at scale, and refining how it evaluates risks of misalignment and shutdown resistance.
The update also requires safety case reviews before higher-risk models can be deployed.
“Our Framework will continue evolving based on new research, stakeholder input and lessons from implementation,” DeepMind said in an announcement. “We remain committed to working collaboratively across industry, academia and government.”
In parallel with its governance work, DeepMind is extending agentic capabilities to robotics. Its Gemini Robotics 1.5 models now integrate reasoning, planning and vision-language-action skills to complete multistep physical tasks, signaling how frontier AI is moving beyond digital assistants to applications in the real world.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
