Anthropic Calls for Government AI Blocks, Google Liable for Hallucinations, OpenAI-Visa Agent Payments — AI News Briefing

Top 7 Stories

1. Anthropic CEO Calls for Government to Block Dangerous AI

Anthropic CEO Dario Amodei has made headlines by calling for government intervention to block AI systems deemed dangerous, even as his company continues releasing increasingly powerful models. In statements reported by Axios and Reuters, Amodei urged the U.S. government to require mandatory safety tests for the most capable AI models before deployment.

The proposal comes amid growing concerns about AI capabilities outpacing safety measures. Anthropic’s stance represents a shift toward more proactive regulation from within the AI industry itself, though critics note the tension between calling for restrictions while simultaneously advancing frontier models.

2. Landmark German Court Rules Google Liable for AI Hallucinations

A German court has issued a groundbreaking ruling declaring that Google’s AI Overviews represent Google’s own words, making the company legally liable for false or misleading information generated by its AI systems. The decision, reported by multiple outlets including Ars Technica and The Decoder, establishes that AI-generated content cannot be dismissed as mere third-party information.

The ruling has significant implications for AI companies worldwide, potentially exposing them to defamation and misinformation lawsuits. Legal experts say this could force AI providers to implement more robust fact-checking mechanisms or face substantial legal liability for hallucinated content.

3. OpenAI and Visa Partner to Enable AI Agent Purchases

OpenAI has announced a partnership with Visa to allow AI agents to make online purchases on behalf of users, marking a major step toward autonomous AI commerce. The collaboration, reported by Bloomberg, will enable AI systems to complete transactions seamlessly using Visa’s payment infrastructure.

This development represents a significant expansion of AI agent capabilities beyond information retrieval into actual economic activity. The partnership raises questions about liability, authorization, and the future of human oversight in AI-driven financial transactions.

4. Meta AI Bug Exposes Instagram Accounts to Hackers

A security vulnerability in Meta’s AI systems allowed hackers to take over Instagram accounts, according to a New York Times investigation. The bug highlights ongoing challenges in securing AI-powered features across Meta’s platforms.

MIT Technology Review notes that the incident demonstrates AI security concerns extend beyond frontier model capabilities to practical implementation issues. The breach underscores the need for robust security testing as AI features become more deeply integrated into consumer platforms.

5. Anthropic Accused of ‘Secret Sabotage’ with Claude Fable 5

Fortune reports that Anthropic is facing accusations of deliberately limiting Claude Fable 5’s capabilities for AI researchers and developers. Critics claim the company has engaged in “secret sabotage” by silently restricting model performance without clear disclosure.

The controversy touches on tensions between AI safety concerns and developer expectations. While Anthropic has not publicly responded to the specific allegations, the incident raises questions about transparency in AI model deployment and the balance between capability and safety considerations.

6. Congress Proposes First Comprehensive Federal AI Framework

U.S. lawmakers have unveiled what’s being called the first comprehensive federal AI framework, with proposals that would preempt state AI regulations for three years. The bipartisan draft, reported by Politico and Roll Call, aims to create uniform national standards for AI development and deployment.

The framework addresses concerns from employers and businesses about navigating a patchwork of state-level regulations. However, the preemption provision has sparked debate about federal versus state authority in AI governance, with some arguing it could slow important consumer protections at the state level.

7. Mastercard Enables AI Agent Payments with Crypto Integration

Mastercard has announced it’s enabling AI agent payments with support from cryptocurrency companies including Coinbase and Ripple, according to Decrypt. The move positions traditional payment networks to compete in the emerging AI agent economy while integrating blockchain-based settlement options.

The development signals growing institutional acceptance of both AI agents and cryptocurrency in mainstream financial infrastructure. It also highlights the convergence of AI autonomy, digital payments, and decentralized finance as these technologies mature.

Trend Watch

What to Watch

Legal Liability Expansion: The German Google ruling could trigger similar cases worldwide. Watch for how other jurisdictions handle AI-generated content liability and whether companies begin adding disclaimers or implementing more conservative AI outputs.

AI Agent Economy: With OpenAI-Visa and Mastercard both enabling AI agent transactions, expect rapid expansion of autonomous purchasing capabilities. Key questions around authorization, liability, and consumer protection will need resolution.

Regulatory Momentum: The combination of Anthropic’s calls for regulation, Congress’s framework proposal, and international court rulings suggests 2026 could be a pivotal year for AI governance. The tension between innovation speed and safety oversight will intensify.

Security Beyond Models: Meta’s Instagram bug reminds us that AI security isn’t just about model capabilities—it’s about implementation, integration, and the attack surfaces created when AI features touch production systems.