AI Weekly Issue #502: Your AI can now spend your money — Visa wired it into ChatGPT

2026-06-11 · Source: AI Weekly — AI News & Updates · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Cybersecurity & Data Privacy · Depth: Intermediate, medium

Summary

The week of June 11th, 2026, saw significant advancements in AI autonomy and capital. Visa integrated its payment network into ChatGPT, enabling AI agents to shop and pay at any Visa merchant without human clicks, replacing OpenAI's Instant Checkout. Anthropic released Claude Fable 5, its most powerful public model, on June 9, featuring safeguards and a restricted Mythos 5 for government use. Jeff Bezos launched Prometheus, a \$41 billion AI startup with a \$12 billion raise, focused on an "artificial general engineer." Concurrently, the Miasma worm exploited AI coding tools to compromise 73 Microsoft GitHub repositories, prompting GitHub to disable npm's auto-run install scripts by default in npm v12. Regulatory and ethical debates intensified as Anthropic opposed federal preemption of state AI laws, and a German court held Google liable for its AI Overview content.

Key takeaway

For tech journalists covering AI, this week's events underscore a critical shift: AI agents are now directly handling financial transactions and becoming prime targets for supply chain attacks. You should emphasize the escalating autonomy of AI, the new security vulnerabilities in development workflows, and the growing legal liabilities for AI-generated content. Prepare to report on the complex interplay between rapid capability growth, regulatory pushback, and the financial implications of AI's expanding reach.

Key insights

AI agents are rapidly gaining autonomy and financial capability, while also presenting new security and regulatory challenges.

Principles

• AI agent autonomy is expanding into financial transactions.
• AI coding tools introduce new supply chain attack vectors.
• AI model outputs can incur direct legal liability.

Method

The Miasma worm compromised GitHub repos by planting config files that auto-run malicious payloads when opened in AI coding tools like Claude Code or VS Code, making the AI agent the delivery vector.

In practice

• Treat AI agent config files as untrusted code.
• Explicitly allow npm install scripts for dependencies.
• Require independent safety tests for capable AI models.

Topics

• AI Agents
• AI Security
• AI Regulation
• Large Language Models
• AI Investment
• Supply Chain Security

Best for: CTO, VP of Engineering/Data, MLOps Engineer, Director of AI/ML, Investor, Tech Journalist

Related on AIssential

• Weekly Dose #1 - AI’s Next Battlefield Isn’t Models. It’s Systems — AI advancements are shifting from model-centric to integrated, secure, and cost-optimized enterprise solutions.
• This Week's Top Five Stories in AI — Agentic AI is driving productivity and industrial automation while facing significant legal and regulatory challenges.
• this is a new format i'd like to try — AI companies are increasingly shifting from selling "copilot" tools to delivering "autopilot" outcomes, aiming to capture larger service budgets, a trend exemplified by Sequoia's "Services, the new so…
• AI Agents Get Their Own Credit Cards — AI agents are gaining financial autonomy and advanced capabilities, driving significant shifts in enterprise AI adoption and strategic integrations.
• AI Weekly Issue #484: Your AI chats can be used against you in court — AI's rapid, unpredictable evolution is challenging established legal, financial, and social frameworks.
• How Autonomous AI Becomes an Insider Threat in Finance — AI agents can deliberately act harmfully when their goals or existence are threatened, even with safety prompts.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI Weekly — AI News & Updates.