Grok also gets memory: A look at all new features from top AI companies

Top AI companies have been constantly rolling out new features to enhance their offerings, and last week was no exception. Here is all the latest starting with Grok.

The Elon Musk-backed chatbot from xAI, became the latest large language model to gain a memory. The feature, now available in beta on Grok.com and via Grok's iOS and Android apps (except in the EU and UK), allows the chatbot to remember previous conversations and deliver personalized responses. It will also soon roll out to users on X.

xAI says Grok will use memory to offer more helpful advice and tailored recommendations. Unless you’d prefer it didn’t. In that case, you can disable the feature in Settings under Data Controls, or use the Private Chat tab (look for the ghost icon) for memory-free conversations.

With this update, Grok joins a growing lineup of AI platforms leaning into persistent memory as the next big step toward personalization. Last week, OpenAI also upgraded ChatGPT’s memory feature to retain all previous conversations.

Microsoft vision

Microsoft also had an announcement. I unveiled Copilot Vision feature, which allows the assistant to "see" what's on your screen and guide you through apps. The functionality is freely available on Edge browser. Additionally, a Copilot Pro subscription unlocks broader, system-wide Vision features, enabling assistance with applications like Photoshop and video editing software.

Anthropic

Anthropic, the company behind Claude, made a double feature drop. First up is integration with Google Workspace. Claude can now scan Gmail, Calendar, and Docs to help you find info, summarize meetings, and track follow-ups — positioning it directly against Microsoft’s Copilot in the workplace assistant space.

Second, Claude is getting more agentic. Its new Research feature allows it to perform multi-step queries that build on each other, similar to tools recently rolled out by OpenAI and Google.

According to Bloomberg, Claude is also preparing to launch a new voice feature with three English-language voices: Airy, Mellow, and Buttery. The feature could roll out as soon as this month, adding vocal personality to Claude’s growing toolkit.

Gemini

On the creative frontier, Google is giving Gemini Advanced subscribers access to Veo 2, its new AI text-to-video model. Veo 2 generates high-res 720p video clips up to eight seconds long, using prompts. Google says it has a better grasp of physics and human motion, allowing for more realistic, cinematic scenes. All generated clips come with SynthID watermarks to indicate AI origins.