Catch up on select AI news and developments from the past two weeks (in no particular order):
OpenAI launches ChatGPT agent to automate real-world tasks. OpenAI has released ChatGPT agent, a new tool that combines capabilities like Web browsing, deep research synthesis, and terminal access to complete complex, real-world tasks such as analyzing competitors or planning events. Available to Pro, Plus, and Team users, it includes integrations with Gmail and GitHub and scores highly on difficult benchmarks. Safety features include biological-threat detection and disabled memory to prevent data leakage. This rollout marks OpenAI’s biggest step toward practical agentic AI that takes autonomous action on users’ behalf. Importance for marketers: Enables automation of research, reporting, and task management, freeing time for strategic decision-making.
Perplexity’s Comet browser redefines productivity with agentic AI. Comet, Perplexity AI’s Chromium-based browser, introduces agentic browsing that allows users to offload tasks like trip planning, restaurant booking, and deal-hunting to an embedded assistant. Users can engage with the AI agent to perform real-time research while continuing other tasks, increasing efficiency. Though still in beta, Comet represents a shift toward AI-integrated Internet navigation. It emphasizes hands-free Web activity and productivity gains, echoing the push for agent-first computing. Importance for marketers: Highlights a new AI touchpoint—agent-powered browsers—that may influence how consumers search, shop, and engage with content.
ChatGPT aims to replace Excel and PowerPoint with native tools. OpenAI is testing ChatGPT features that allow users to create and edit spreadsheets and presentations directly, without Microsoft Office. Compatible with .xlsx and .pptx formats, these tools could position ChatGPT as a serious productivity suite rival. OpenAI is also expanding its agent capabilities to automate full workflows like report compilation or appointment booking. Despite its Microsoft partnership, OpenAI’s development of overlapping features indicates strategic divergence. Importance for marketers: Offers a streamlined alternative for campaign planning and reporting tasks, potentially lowering software costs and simplifying tool stacks.
US companies pledge $90B in AI and energy investments at Trump summit. At a summit in Pennsylvania, major US firms, including Google, Meta, Blackstone, and CoreWeave, announced investments totaling $90 billion across AI infrastructure and energy. Google is investing $25B in data centers and partnering on hydropower. Meta is building massive AI data centers, and CoreWeave is developing a $6B AI facility. First Energy, Constellation Energy, and others are expanding grid capabilities. These investments reflect the growing infrastructure demands of AI at scale. Importance for marketers: Indicates continued AI investment momentum that will expand cloud access, lower latency, and support increasingly sophisticated marketing technologies.
Google adds AI business calling and Gemini 2.5 Pro to search. Google has launched AI-powered business-calling, allowing users to request appointment and pricing info from local businesses without making a call themselves. It also upgraded Search’s AI Mode with Gemini 2.5 Pro for advanced reasoning, and it added “Deep Search” to generate cited reports from multiple sources. These features expand the utility of Search from information retrieval to research assistant. Importance for marketers: Marketers should expect shifts in search behavior and local SEO impact, as AI intermediaries begin to dominate decision-making touchpoints.
Mistral debuts Voxtral, an open-source AI audio model. French AI startup Mistral has launched Voxtral, an open-weight speech model for transcription, summarization, and command execution. Voxtral supports multiple languages and comes in high-parameter and lightweight versions, with pricing aimed at undercutting OpenAI’s Whisper. It integrates with Mistral’s chatbot and offers API access for production use. Voxtral is Mistral’s entry into audio AI, following its text-based Magistral reasoning models. Importance for marketers: Presents a low-cost alternative for multilingual audio-to-text workflows, useful for video content, podcast editing, and accessibility efforts.
Amazon launches AgentCore to build custom AI agents for work. AWS has introduced AgentCore, a platform for building interconnected AI agents to handle data analysis, code generation, and task automation. Compatible with multiple models and frameworks, AgentCore includes a dashboard for performance tracking and a marketplace for sharing agents. Although it’s enterprise-ready, adoption may lag without training and internal buy-in. Importance for marketers: Empowers marketing teams to build tailored agents for campaign optimization, performance monitoring, or content automation—if organizational support and education are in place.
Samsung explores smart AI wearables beyond glasses. Samsung is exploring AI-powered wearable devices like smart earrings, rings, and necklaces as companions to smartphones. The move aligns with industry efforts to create less obtrusive, hands-free interfaces. Smart glasses remain a near-term focus, but Samsung’s broader vision hints at future product categories. Past efforts like Humane’s AI Pin and Rabbit R1 struggled, but Samsung is taking a more conservative, companion-based approach. Importance for marketers: Marketers may need to rethink user experience design and content strategies as wearable AI interfaces evolve.
Germany unveils AI strategy to boost economic output by 2030. Germany’s new AI strategy targets 10% of GDP to be AI-driven by 2030, with plans for EU-funded AI ‘gigafactories’ by 2027, quantum computing initiatives, and enhanced commercialization of research. The plan also calls for better AI adoption in industry and administration. Importance for marketers: Reflects the European Union’s intent to remain competitive in AI, which could affect regulatory frameworks, funding access, and the AI vendor landscape in EMEA.
OpenAI developing in-chat checkout system for ChatGPT. OpenAI is testing a payment checkout feature inside ChatGPT that would let users buy products directly through the chatbot. Retail partners like Shopify are already previewing the system, which would enable OpenAI to earn a commission on sales. This represents a potential new revenue stream beyond subscriptions. Importance for marketers: Opens a new commerce channel and could evolve ChatGPT into a full-funnel sales assistant—critical for brands investing in conversational commerce.
Microsoft’s Copilot Vision now sees entire screen for context-aware help. A new Copilot Vision update lets the AI tool analyze full-screen content on Windows, not just limited app views. Users can enable it to provide coaching, resume tips, or content suggestions based on what’s visible. This differs from Microsoft Recall in that it’s user-triggered and not continuously recording. Importance for marketers: Raises expectations for productivity AI to offer real-time, visual-context suggestions—useful for marketers reviewing creative assets or refining messaging.
Study finds humans adopting ChatGPT-style vocabulary in speech. Researchers found increased use of “GPT words” like “delve,” “underscore,” and “meticulous” in YouTube videos and podcasts post-ChatGPT. The study, still under peer review, suggests that cultural feedback loops are forming as AI influences human communication patterns. Importance for marketers: Offers insight into evolving language norms, which can affect copywriting tone, brand voice, and content resonance with audiences.
AI experts call for ‘chain-of-thought’ monitoring to improve model safety. Researchers from OpenAI, DeepMind, and others urge the industry to monitor how AI models reason via “chains of thought” (CoT) to improve transparency and safety. CoT visibility may be key to understanding alignment in AI agents, but could diminish as models evolve. Importance for marketers: Transparency in reasoning will be essential when using AI for decision-support in regulated sectors like finance, healthcare, or legal marketing.
Claude now supports Canva workflows for design tasks. Claude AI users can now generate, edit, and summarize Canva projects through natural language prompts, using the Model Context Protocol (MCP) standard. The integration eliminates the need to switch apps for design tasks. Other MCP integrations with Figma, Stripe, and Notion are also supported. Importance for marketers: Simplifies creative production workflows by embedding design control into AI chats—ideal for lean teams managing social content or presentations.
Judge allows US authors to form class action against Anthropic. A federal judge ruled that authors can proceed with a class-action lawsuit against Anthropic, alleging it illegally downloaded millions of pirated books to train AI models. The ruling could expose Anthropic to billions in liability. Importance for marketers: Reinforces the legal risk of using generative AI trained on unlicensed data—relevant for brand teams building content pipelines with third-party models.
UN urges stronger safeguards against AI-driven deepfakes. The UN’s International Telecommunication Union (ITU) has called for robust standards and detection tools to combat AI-generated deepfakes, citing risks to elections, financial security, and public trust. The report, released at the AI for Good Summit, recommends content authentication measures and watermarking standards for video—currently 80% of internet traffic. Experts from Adobe, Digital Medusa, and others stressed the need for global coordination and user education. Importance for marketers: Highlights the growing pressure for transparency and verification in visual content—critical for brand integrity and consumer trust in AI-enhanced campaigns.
You can find the previous issue of AI Update here.
Editor’s
note: GPT-4o was used to help compile this issue of AI
Update.
