Amazon Triggers Anthropic Ban, Meta Unwinds Manus Deal, OpenAI Probed — AI News Briefing

Top 7 Stories

1. Amazon Security Research Triggers White House Order to Ban Anthropic’s Fable 5 and Mythos 5

Amazon’s internal cybersecurity research has reportedly been the catalyst for the U.S. government’s unprecedented order to Anthropic to block all foreign access to its most powerful models, Claude Fable 5 and Claude Mythos 5. According to the Wall Street Journal, Amazon produced a paper claiming it could extract cyberattack-useful information from Fable 5 through a series of prompts, and CEO Andy Jassy shared these findings directly with White House officials.

The directive forced Anthropic to cut off access entirely — for all customers worldwide, including its own employees — rather than implement geo-restrictions. Anthropic pushed back publicly, stating it “disagrees that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people.” The episode marks the first time the U.S. has used export control authority to restrict a frontier AI model, setting a precedent that treats AI capabilities similarly to advanced chip exports.

2. Meta Begins Dismantling $2 Billion Manus Acquisition After Beijing Orders Reversal

Meta has started unwinding its $2 billion acquisition of AI agent startup Manus after Beijing ordered the deal reversed, TechCrunch reports. The move highlights the intensifying geopolitical tug-of-war over AI talent and technology between the U.S. and China.

Manus, known for its autonomous AI agent capabilities, had become a flashpoint in the growing friction around cross-border AI deals. Beijing’s intervention signals that China is increasingly willing to block the transfer of AI-related intellectual property and talent to American tech giants, even after deals have been completed. For Meta, the reversal represents a significant strategic setback in its AI agent ambitions.

3. OpenAI Faces Investigation from Multiple State Attorneys General

OpenAI is under investigation by a coalition of state attorneys general, with probes covering everything from the company’s advertising policies to its handling of health-related user data. While the specific states involved have not been publicly confirmed, the scope of the inquiry suggests a broad regulatory push into how AI companies operate in consumer-facing markets.

The investigation arrives at a critical time for OpenAI as it transitions toward a for-profit structure and expands its consumer products. Questions about data practices, advertising transparency, and user safety could shape the regulatory framework that will govern AI companies for years to come.

4. KPMG Pulls Report on AI Usage After Apparent Hallucinations

Big Four accounting firm KPMG has been forced to retract a published report on AI usage after readers identified apparent hallucinations — fabricated statistics and sources — within the document. The incident, covered by TechCrunch, underscores a growing irony: AI tools used to research AI are producing unreliable results.

The retraction raises questions about quality control processes at major professional services firms that are increasingly relying on AI-generated content for client-facing research. It also highlights the persistent challenge of AI hallucination in high-stakes business contexts where accuracy is paramount.

5. India Debates AI Self-Reliance After Anthropic Access Suspension

The Anthropic model shutdown has ignited a fierce debate in India about the country’s dependence on foreign AI providers. Indian tech leaders are calling the episode a “wake-up call” for the nation’s AI ambitions, with many arguing that reliance on U.S.-based frontier models leaves Indian businesses and researchers vulnerable to geopolitical decisions made in Washington.

The discussion touches on broader themes of digital sovereignty and the need for India to accelerate development of domestic foundation models. Several Indian AI startups have already seen surges in interest from enterprise customers seeking alternatives that won’t be subject to sudden access restrictions.

6. Meta’s AI Unit Engineers Report “Soul-Crushing” Working Conditions

A new report from TechCrunch describes Meta’s dedicated AI unit — a 6,500-person division central to the company’s AI strategy — as a “soul-crushing gulag” according to engineers working inside it. Sources say the unit is on the verge of internal revolt due to intense working conditions, unclear priorities, and management pressure.

The revelations come as Meta faces mounting competition from OpenAI, Anthropic, and Google in the frontier AI race. If the reports are accurate, internal morale issues could hamper Meta’s ability to execute on its ambitious AI roadmap, which includes major investments in open-source models and AI-powered social features.

7. Apple Unveils AI Photo Editing Tools in iOS 27 at WWDC 2026

Apple has introduced its first serious native AI photo editing features in iOS 27, including AI-powered reframe, extend, and clean-up tools announced at WWDC 2026. The Verge reports that while the features are relatively tame compared to Google Pixel’s AI editing capabilities, they represent a major shift for the iPhone’s native Photos app.

The tools allow users to intelligently extend image borders, remove unwanted objects, and reframe compositions — all processed on-device. The launch signals Apple’s gradual embrace of generative AI in consumer-facing products, following its Gemini-powered Siri overhaul also announced at WWDC.

Trend Watch

What to Watch

• Anthropic’s next move: Will Anthropic pursue legal action against the government order, or accept the new reality of AI export controls? The company’s response will define how AI firms navigate national security directives.
• Meta-Manus fallout: Watch for regulatory responses in both Washington and Beijing. This deal’s collapse could chill future cross-border AI acquisitions.
• OpenAI investigation scope: If state AGs gain access to OpenAI’s internal data practices, the findings could reshape how all AI companies handle user information.
• Frontier model access as geopolitical tool: The Anthropic precedent means AI model access is now explicitly tied to national security — expect similar moves from other governments.