OpenAI 'Operator' 🤖, Anthropic Citations API 📚, OpenAI Operator Data Retention 🔒

In partnership with

Start learning AI in 2025

Everyone talks about AI, but no one has the time to learn it. So, we found the easiest way to learn AI in as little time as possible: The Rundown AI.

It's a free AI newsletter that keeps you up-to-date on the latest AI news, and teaches you how to apply it in just 5 minutes a day.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses – tailored to your needs.

Sign up to start learning.

💬 OpenAI Developments

Open AI's browser-enabled AI agent now released.

OpenAI has launched "Operator," an AI agent capable of autonomously performing web-based tasks such as booking trips and purchasing groceries. Initially available to ChatGPT Pro users in the U.S., Operator interacts with web pages by typing, clicking, and scrolling, aiming to enhance productivity by automating routine online activities.

Tiktok's parent company challenges Open AI with the Doubao model

ByteDance, the parent company of TikTok, has released an updated AI model named Doubao-1.5-pro, aiming to challenge OpenAI's latest reasoning models. This move is part of a broader effort by Chinese companies to advance in AI reasoning and challenge global competitors like OpenAI, backed by Microsoft.

Stargate artificial intelligence project to exclusively serve OpenAI

Stargate, an AI initiative supported by SoftBank, has secured an exclusive deal to provide OpenAI with advanced computing resources and infrastructure. This partnership highlights SoftBank's ambitions in the AI industry, positioning Stargate as a key player in supporting OpenAI’s expanding capabilities, while emphasizing confidence in the U.S. as a leader in AI innovation.

🌐 AI in Various Sectors

Healthcare: Yidu Tech showcased its AI-driven healthcare innovations at Winter Davos 2025, detailing global expansion plans and successes in digital health, clinical trials, and insurance solutions. Founder Ms. Gong Rujing highlighted the company’s resilience, collaborative approach, and commitment to making precision healthcare accessible worldwide.

Robotics: Yann LeCun, Meta's chief AI scientist, anticipates a significant shift in AI architectures within the next three to five years, moving beyond current models like ChatGPT. He also envisions the upcoming decade as transformative for robotics, with AI advancements leading to more autonomous and adaptable machines.

Social Media and Interaction: Instagram is adopting a vertical post format, expanded Reel run times, and a host of profile layout tweaks—echoing TikTok’s design and features. These updates follow TikTok’s ban in the U.S., with Instagram seizing the opportunity to attract users seeking a similar vertical video experience.

🚀 Tech Industry Moves

Anthropic has launched a Citations feature for its Claude AI platform, enabling developers to link AI-generated answers to specific source documents like emails, knowledge bases, or other reference materials. This tool is designed to combat AI hallucinations by grounding responses in verifiable information, improving accuracy, transparency, and reliability for applications like customer support and content generation.

Google has expanded its Gemini AI's smart home control features to all users, enabling natural language management of connected devices such as lights, thermostats, and speakers through the Google Home extension. This update allows for commands like adjusting lighting based on room conditions or executing multiple actions simultaneously, enhancing the integration and convenience of smart home management.

Google AI has introduced "Learn-by-Interact," a data-centric framework designed to enhance the development of adaptive and efficient large language model (LLM) agents. This framework automates the synthesis of interaction data by leveraging accessible resources such as documentation and tutorials, thereby improving the adaptability and efficiency of LLM agents.

NVIDIA has unveiled a series of agentic AI innovations aimed at transforming enterprise technology. These include new blueprints for building AI agents, a specific blueprint for video analysis, and the "Llama Nemotron" suite of open large language models, all designed to tackle real-world challenges and unlock new opportunities in various industries.

Perplexity AI, backed by Nvidia and Jeff Bezos, has launched "Perplexity Assistant," an AI-powered tool for Android devices designed to compete with existing assistants like OpenAI's ChatGPT, Apple's Siri, and Amazon's Alexa. This assistant can perform tasks such as booking dinner reservations, hailing rides, and setting reminders, and is available in 15 languages on the Google Play Store.

⚖️ AI Ethics and Regulation

OpenAI’s new AI agent, Operator, may store users’ chats and screenshots for up to 90 days—even after they’re manually deleted—compared to 30 days for ChatGPT. The company cites abuse prevention as the reason for the longer retention period, which allows it to investigate potential misuse. While Operator captures screenshots to inform its browsing and task automation, OpenAI says it doesn’t store passwords or screenshots during “take over” mode and limits data access to authorized personnel and trusted service providers.

💡 Emerging AI Technologies and Companies

People are benchmarking AI by having it make balls bounce in rotating shapes

Informal tests on X have AI models coding a bouncing ball constrained by a rotating shape, with varying success—a simple but telling challenge for logic, physics simulation, and collision detection. Some models, like Chinese lab DeepSeek’s R1, outperform premium options such as OpenAI’s o1, illustrating how ad hoc tests often highlight AI’s inconsistencies in reasoning and code generation. Yet, these viral benchmarks remain anecdotal measures rather than rigorous, universal standards, underscoring the difficulty of comparing AI performance across diverse tasks.

Adobe Premiere Pro now lets you find video clips by describing them

Adobe has updated Premiere Pro with an AI-powered search feature that recognizes objects, locations, and more within video clips, letting users find footage by simply describing it (e.g., “someone skating with a lens flare”). The media intelligence runs on-device to protect privacy and can also tap into transcript or metadata details. Premiere Pro now supports multi-language caption translation for editing multiple caption tracks simultaneously. After Effects introduces HDR monitoring and a new caching system that uses both RAM and disk storage to speed up playback. Additionally, Canon’s C80 and C400 cameras have added Frame.io’s Camera to Cloud integration via a December firmware update.

🎨 AI in Creative Industries

Sam Altman’s World Project now wants to link AI agents to your digital identity

Sam Altman’s World project (formerly Worldcoin) aims to create tools that verify both humans and their AI agents online by scanning people’s irises and giving them unique IDs on the blockchain. This could let AI agents act on users’ behalf with proof they’re linked to a real person, addressing concerns like website security and bot misuse. World’s approach, integrated with Altman’s other AI-centric efforts, could significantly reshape how digital identity and AI intersect, allowing businesses to selectively permit verified AI agents while preventing malicious bots.

🌍 Global AI Race

ITU Launches AI Skills Coalition with AWS, Microsoft, and Partners to Address Global AI Training Gap

The International Telecommunication Union (ITU) has launched the AI Skills Coalition, partnering with over 25 organizations, including Amazon Web Services, Microsoft, and Cognizant, to bridge the global AI skills gap. This initiative aims to provide accessible AI education and training, focusing on underserved communities, and will offer resources such as self-paced courses, webinars, and workshops to promote inclusive participation in the AI revolution.

If you want more updates related to AI, subscribe to our Newsletter