AI news roundup, April 2026: Razorpay launches agentic payments, SpaceX partners with Cursor, National Security Agency works with Anthropic, Amex introduces agentic commerce stack, while Google expands Gemini, and OpenAI attracts top talent.
Razorpay launches “Agentic Payments” stack
India-based fintech Razorpay introduced an Agentic Payments stack today. It is designed to let AI agents initiate and complete transactions autonomously, recover lost revenue, automatically manage disputes, and create ChatGPT storefronts. The merchant suite update builds on the firm’s earlier conversational payments experiments with UPI and AI interfaces, including recent signals around deeper integrations with platforms like Google Pay.
Razorpay is positioning itself for the next phase of fintech, where AI agents, not users, drive transactions, reducing friction in checkout and enabling automated commerce flows. In the U.S., almost 40% of consumers use AI tools for product discovery, and about a quarter of shoppers have already tried buying something using agentic AI interfaces.
SpaceX signs major deal with AI coding platform Cursor
SpaceX has signed a strategic agreement with Cursor, an AI-native code editor used by 67% of Fortune 500 companies, with terms reportedly allowing for a $60 billion acquisition option. The partners are also planning to leverage xAI’s Colossus infrastructure to scale up Cursor’s agentic coding model Composer.
This way, SpaceX is strengthening its AI capabilities by integrating advanced coding automation into its ecosystem, potentially aiming to accelerate software development across aerospace and its affiliated AI initiatives.
Surge in AI upskilling for green jobs
New industry reporting in April 2026 highlights a growing push among companies to train workers in AI skills tailored to sustainability roles, including energy optimization and climate analytics. Businesses face increasing pressure to meet climate targets while improving efficiency, making AI-driven sustainability expertise a critical workforce priority.
The fastest growing skills among green talent are:
- responsible AI (up 517% YoY),
- LangChain (up 379% YoY),
- large language models (up 219% YoY),
- prompt engineering (up 197% YoY),
- chatbot development (up 171% YoY), and
- ChatGPT (up 171% YoY).
Using the new AI skills might be highly beneficial for green talent, too. According to the fresh Gensler workplace trend report, people who regularly use AI in both their work and personal lives spend more of their workweek learning and collaborating virtually compared to those who use AI less frequently. Namely, AI power users spend 1.5x as much time learning as later adopters. And despite spending less time collaborating in person, AI power users report stronger team relationships, greater encouragement to share new ideas, and more meaningful friendships at work, pointing to new patterns of connection enabled by technology.
NSA partners with Anthropic on AI for national security
The National Security Agency is working with Anthropic to deploy advanced AI systems in defense environments. The controversial nature of this collaboration is that the agency is reportedly using Anthropic Mythos Preview, the model that was not launched publicly due to its high capability to discover cybersecurity vulnerabilities. All this is happening just weeks after the NSA’s parent agency, the Department of Defense, labeled Anthropic tools a “supply-chain risk” that can threaten national security, since the company refused to allow Pentagon officials unrestricted access to its model’s full capabilities.
Though the specifics of the NSA collaboration remain to be clarified, the U.K.’s AI Security Institute also confirmed access to Mythos. For once, governments are accelerating the adoption of large language models for intelligence analysis and operational support, aiming to improve decision-making speed and accuracy in national security contexts. The second point is that this particular AI model has great potential for both boosting and destroying cybersecurity at the enterprise and national levels.
As Nvidia CEO Jensen Huang warned, China already has the computing power and data center capacity necessary to train an AI model similar to the level of Anthropic’s AI model Claude Mythos, so Western economies try to maintain the cybersecurity edge, accelerating their cooperation with private AI companies. Banks are also encouraged to deploy Mythos proactively for self-auditing.
Meanwhile, Anthropic’s latest publicly available model, Claude Opus 4.7 debuted on Amazon Bedrock next-generation inference engine for generative AI inferencing and fine-tuning workloads.
Google rolls out “personal intelligence” and enterprise agentic stack upgrades in Gemini
Google announced new updates to its Gemini app, focusing on “personal intelligence” — AI that adapts to user behavior across tasks and devices. The update also emphasizes more on-device processing. For individual users, the Gemini Nano Banana 2 model now helps create personalized images using their own interests and photos. Meanwhile, enterprises get the chance to build their agentic task force.
With these updates, Google is competing to make its AI assistants more context-aware and persistent, strengthening user engagement while improving privacy through local processing.
Talent shift: Hiro joins OpenAI ecosystem
AI developer Hiro is joining OpenAI, continuing the trend of high-profile talent moving into top AI labs. Leading AI firms are consolidating expertise to accelerate development of advanced models and agent-based systems, where talent concentration directly impacts innovation speed. Since 2023, OpenAI has lured in some researchers earlier involved in large language model training and reinforcement learning at Google’s DeepMind, and engineers from Meta’s LLaMA and large-scale training infrastructure teams. Meanwhile, many former OpenAI employees moved on to form their own AI startups and projects.
Amex launches Agentic Commerce Experiences (ACE) Developer Kit, takes liability for AI-driven transactions
American Express has introduced its Agentic Commerce Experiences (ACE) Developer Kit, formalizing AI-agent-initiated payments in production environments. Furthermore, Amex is explicitly accepting liability for transactions executed by machine-based “customers” (AI agents) when they operate within approved frameworks and authentication flows.
By owning liability, Amex is effectively enabling merchants and developers to treat agent-driven transactions as legally and financially equivalent to human card payments, but with added verification layers and governance controls.
The ACE Developer Kit provides the tooling for developers to integrate agentic payment flows, including:
- authentication for AI agents acting on behalf of users;
- transaction authorization frameworks for machine-initiated purchases;
- compliance and risk controls adapted for non-human “customers.”
