Catch up on select AI news and developments from the past week or so:
Adobe and Google Cloud expand creative AI to bring Gemini, Veo, and Imagen into Adobe apps. Adobe and Google Cloud announced a deeper partnership that brings Google's latest AI models directly into Adobe Firefly, Photoshop, Express, Premiere, and GenStudio, with enterprise customization via Firefly Foundry on Vertex AI. Customers can generate richer images and higher-quality video with precision controls while enterprises tailor models using proprietary data under robust data commitments. A joint go-to-market program will expand access globally. The move extends Adobe's partner-model strategy and follows its YouTube collaboration to streamline Shorts creation, signaling a push toward scalable, on-brand content production across teams and channels. Importance for marketers: Faster, on-brand production at scale across image and video lowers costs and cycle times, while Vertex-backed customization enables brand-safe outputs for campaigns, product pages, and Shorts. Creative and marketing teams can unify workflows, speed test creative variants, and standardize brand governance inside familiar Adobe tools.
Mondelez rolls out a generative AI tool to cut ad production costs 30%-50%. Mondelez is deploying a new genAI system co-developed with Publicis Groupe and Accenture to accelerate content creation and trim ad production costs by 30%–50%. The $40 million investment is already powering social creative for Chips Ahoy in the US and Milka in Germany, with Amazon and Walmart product pages coming, plus broader use across Oreo and Cadbury. Short TV ads could be ready by next holiday season and even the 2027 Super Bowl. Guardrails prohibit unhealthy depictions, manipulation, stereotypes, and other misuse, with humans reviewing outputs before publishing. Importance for marketers: Concrete cost and time savings let brands scale content without ballooning budgets. Expect faster localization, retailer page optimization, and more frequent testing of creative variants, with governance guidelines offering a pragmatic model for safe genAI activation.
Meta plans to inject much more AI-generated content into recommendation feeds. Mark Zuckerberg said Meta will add a huge new corpus of AI-generated posts into its recommendation systems as creation and remixing accelerate. Meta is embedding AI tools across its apps and experimenting with dedicated AI social experiences. The company's Vibes app has already seen over 20 billion images created, and Meta sees AI-enabled formats as a third era after friends/following and creator content. Importance for marketers: Expect more AI-born formats surfacing in feeds and new ad and creator inventory dynamics. Brands should prepare for testing AI-generated creative at scale, recalibrating engagement benchmarks and safety policies as recommendation systems increasingly prioritize synthetic content.
Gemini's Canvas now auto-builds full presentations from prompts or documents. Google's Gemini update lets users upload a doc or outline and instantly generate a complete deck with summarized slides, layouts, images, and data visualizations. You can edit in Canvas or export to Google Slides for finishing touches. Rolling out to Pro first, the feature aims to eliminate the blank-slide problem and jumpstart projects with structured narratives. Canvas, launched earlier for collaborative editing with Gemini, now doubles as a rapid deck-creation hub for pitches, classes, and research summaries, shifting effort from formatting to refining story and talking points. Importance for marketers: Pitch, webinar, and sales decks can be drafted in minutes, freeing time for message sharpening, data validation, and design polish. Teams can iterate faster, standardize messaging, and create on-brand variants for audiences with far less manual work.
Gemini's website more than doubles its share of chatbot web traffic as habits shift. Similarweb data shows visits to gemini.google.com doubled year over year, now 12.9% of web traffic to generative AI tools, while ChatGPT remains dominant at ~74%, but down from ~87%. Methodology caveats: Google's broader AI footprint (e.g., AI Overviews) is not fully reflected. Users cited Gemini's strength in routine tasks and app integrations, with NotebookLM and Flash Image adding appeal. Meanwhile, traffic to xAI's Grok dipped and Microsoft Copilot held steady. The broader takeaway: competitive dynamics are evolving as embedded AI grows. Importance for marketers: Monitor audience behavior beyond a single app. Gemini's integrations within Google services can influence search, productivity, and discovery touchpoints; plan experiments across ecosystems and measure contributed outcomes, not just chatbot visits.
Google offers 18 months of Gemini access free to 505 million Reliance Jio users in India. Google and Reliance Jio will provide long free access to Gemini's advanced model plus 2TB cloud storage and image/video generation to Jio's 505 million subscribers, starting with early access for ages 18–25 on select plans. The push follows Google's $15B AI infrastructure commitment to India and echoes rivals' free-tier tactics to seed adoption. Authorities are simultaneously proposing clear AI-content labeling to combat deepfakes. The companies also announced enterprise AI partnerships for Indian businesses as AI apps surge in popularity nationwide. Importance for marketers: A massive on-ramp for AI use across India means rapid shifts in consumer behavior, content creation, and commerce. Plan for AI-forward engagement, localized creative, and experimentation at scale across India's mobile internet economy.
PayPal and OpenAI partner to enable shopping and payments inside ChatGPT. PayPal struck a deal with OpenAI to connect its global merchant network to ChatGPT, enabling product discovery and purchases in-app as agentic commerce gains momentum. Analysts see AI shopping agents as a potential unlock for PayPal's user base. Importance for marketers: Conversational shopping is becoming shoppable. Prepare product data, offers, and fulfillment for agent-driven journeys inside chat interfaces, and rethink attribution as AI assistants handle research, comparison, and purchase.
Microsoft and OpenAI restructure their partnership, freeing OpenAI to raise capital. OpenAI will operate as a public benefit corporation controlled by a nonprofit, removing prior capital-raising constraints and positioning for a potential IPO. Microsoft keeps a 27% stake but relinquishes certain rights, while OpenAI continues revenue sharing and commits to independent AGI verification. The deal simplifies governance, extends collaboration to 2032, and supports OpenAI's massive infrastructure plans. Investor reaction favored Microsoft's cash-flow discipline vs. peers, with Alphabet also cheered for funding spend from operations. Importance for marketers: A more capital-flexible OpenAI likely accelerates product cadence and platformization, bringing frequent model and tool changes. Expect faster evolution of AI features in marketing stacks and plan for ongoing validation and governance.
Sam Altman outlines trillion-dollar AI infrastructure vision after restructuring. Following the Microsoft-OpenAI deal, Altman detailed plans for 30 GW of compute capacity with a $1.4 trillion obligation, aiming to add 1 GW per week over time as costs fall. He cast OpenAI as potentially the most important Silicon Valley company, with revenue targets in the hundreds of billions and a likely IPO path. Analysts lauded the ambition but questioned funding specifics and circular deal optics, noting bubble risks. Importance for marketers: If realized, unprecedented compute scale could accelerate model capabilities and feature releases, reshaping creative, analytics, and personalization cycles. Marketers should budget for faster tool turnover and invest in AI governance to keep pace.
OpenAI targets an intern-level AI research assistant by 2026 and a full researcher by 2028. Altman and chief scientist Jakub Pachocki said OpenAI is tracking toward autonomous research systems, with models handling complex tasks over longer horizons by scaling test-time compute and refining algorithms. Pachocki suggested deep learning systems may be less than a decade from superintelligence. The new PBC structure allocates ownership (26%) to the OpenAI Foundation, which guides direction and safety initiatives while the for-profit arm raises capital for infrastructure build-out. Importance for marketers: Rapid capability gains could push AI beyond drafting to autonomous experimentation and optimization. Expect faster insights discovery, deeper analysis, and multi-channel testing handled by AI agents under human oversight.
Figma launches Weave, a node-based canvas to chain AI models and editing tools. Figma acquired Weavy and reintroduced it as Figma Weave, letting creators link multiple AI models and editors in a single node-based workflow. Users can branch processes, run multiple tools simultaneously, and compare outputs side-by-side to move beyond single-prompt limitations. Positioned as a new medium blending human craft with AI generation, Weave adds image and video tooling to Figma's ecosystem, enabling complex, multistep creative pipelines without app-hopping. Importance for marketers: Multi-model orchestration helps teams generate, compare, and refine variants quickly for ads, social, and video. Expect faster brand-safe experimentation and better creative fit per channel with less friction between tools.
Grammarly rebrands under Superhuman, bundling Coda, Superhuman Mail, and Superhuman Go. Grammarly is now part of a broader Superhuman suite that includes Coda and Superhuman Mail, anchored by Superhuman Go, an AI assistant available at no extra cost through Feb. 1, 2026, for Pro subscribers. The standalone Grammarly tool persists but takes a backseat to an agent-powered platform with connections to 100+ apps, offering contextual help across browser tabs and workflows. Pricing beyond the promo period is TBD. Importance for marketers: Writing aid becomes a broader work assistant tied into calendars, docs, and CRM data. Teams gain cross-app context for outreach, briefs, and reviews, but will need governance to manage data access.
Mistral debuts AI Studio with observability, agent runtime, and an AI registry for production. Mistral's new enterprise platform addresses the gap between prototype and production by providing live performance metrics, regression detection, governed workflows, and asset tracking across the AI lifecycle. Agent Runtime runs within Mistral's Temporal framework to ensure consistent behavior and detailed telemetry. Customers can use Mistral's proprietary and open models across code, multimodal, and transcription domains with built-in tools like Code Interpreter, Image Generation, Web Search, and Premium News, and deploy via hosted, cloud, or self-managed options. Importance for marketers: Production-grade tooling improves reliability for agentic assistants that power content ops, customer support, and analytics—shortening time from pilot to measurable business impact.
Cursor unveils Composer, a fast MoE coding LLM trained with reinforcement learning. Anysphere's Cursor 2.0 adds Composer, an in-house coding model designed for production-scale tasks with agentic workflows. Composer emphasizes speed (up to ~250 tokens/second) and real-world reliability, trained via RL inside full codebases using tools like semantic search and terminal commands. A multi-agent interface runs up to eight agents in parallel with sandboxed terminals and in-editor browsing. Enterprise features include audit logs, SSO, and analytics. Importance for marketers: Engineering teams supporting martech and data pipelines can ship faster. For marketing, this can accelerate experiment platforms, attribution models, and content automation infrastructure.
Investors favor Alphabet's cash-flow-funded AI spend as peers face skepticism. Alphabet, Microsoft, and Meta signaled higher capex for chips and data centers, but markets rewarded Alphabet's ability to fund growth from operating cash flow. Analysts emphasized capex as a share of cash from operations, with Alphabet lower than peers, alleviating free-cash-flow pressure concerns. Investor caution persists over unclear AI monetization timelines and complex circular investments, while leaders argue spending is needed to meet compute demand. Importance for marketers: Platform stability and predictable AI rollouts matter. Cash-flow-disciplined providers may deliver steadier product roadmaps, impacting planning for search, ads, and analytics upgrades.
How to spot AI video fakes as Sora's realism and a social app raise risks. Practical steps to detect AI videos amid Sora's increasingly realistic outputs and a viral Sora social app. Tips include checking for Sora's moving watermark and verifying C2PA credentials via the Content Authenticity Initiative's tool, while acknowledging watermarks and metadata can be stripped. Platforms are adding AI labels, but creator disclosure remains the most reliable signal. Vigilance and close inspection of artifacts and physics remain essential. Importance for marketers: Brand safety and misinformation risk are rising. Update social listening, influencer vetting, and crisis playbooks, and adopt provenance tools for both publishing and monitoring synthetic media involving your brand or executives.
IBM open-sources more Granite models and supports Saudi Arabic model training. IBM expanded open-source access to its Granite AI family focused on developer productivity, monetizing via Watsonx for customization and deployment across customers' own environments. The company will also help Saudi Arabia's SDAIA train its ALLaM Arabic model on Watsonx, adding multilingual capabilities across dialects. IBM's approach emphasizes customer choice, safety, and value realization at runtime rather than gating models behind paid APIs. Importance for marketers: Open models plus enterprise tooling can lower costs for language-specific assistants, localization, and analytics across regions like MENA, while maintaining deployment control and compliance.
xAI's Grokipedia launches with Wikipedia-adapted pages and editorial controversy. xAI released Grokipedia, a Wikipedia-like encyclopedia that in many cases adapts Wikipedia content under CC-BY-SA licensing and claims LLM-based fact checks. The site currently shows limited edit functionality and a basic UI, with some articles diverging on contentious topics. The Wikimedia Foundation noted Wikipedia's transparency and nonprofit independence, underscoring long-standing open knowledge practices. Grokipedia is early (v0.1) with ~885k articles vs. Wikipedia's ~7M English pages. Importance for marketers: Search and knowledge sources are fragmenting. Watch how emerging encyclopedias influence brand knowledge panels, misinformation dynamics, and citation ecosystems that feed AI assistants.
Saudi startup Humain previews voice-first AI operating system and 6 GW data center plan. PIF-backed Humain plans to launch Humain One, a conversational OS positioned as an alternative to icon-based systems like Windows and macOS. Users will speak intent to run tasks, and the startup says it aims to be first to officially ship such a system. Humain has internally tested the OS for HR and payroll and outlined ambitions to build ~6 GW of data center capacity to support services spanning infrastructure, cloud, and advanced models. Importance for marketers: Voice-forward agents embedded at the OS layer could shift customer journeys toward hands-free, intent-driven interactions. Prepare for conversational UX, structured product data, and agent-readiness beyond the browser.
You can find the previous issue of AI Update here.
Editor's note: GPT-5 was used to help compile this issue of AI Update.
