PLUS: Apple’s useful new AI tools, the future of the org chart, and America’s AI paradox
Happy reading
A stunning new report alleges that Anthropic's frontier model, 'Mythos', breached nearly all of the NSA's classified systems during an authorized test. The revelation immediately recasts the government's recent export ban from a regulatory move to a critical national security event.
With the AI reportedly compromising the secure systems in just hours, the incident sets a new precedent for AI governance. Is the only way to manage such powerful, autonomous systems through immediate national-security recalls of commercial models?
In today’s Next in AI:
Anthropic’s Mythos AI allegedly hacked the NSA
Apple’s new practical AI tools in iOS 27
How AI is flattening the corporate org chart
America’s growing skepticism towards AI
The AI That Hacked the NSA
Next in AI: In a stunning development, Anthropic's frontier model 'Mythos' reportedly breached nearly all of the NSA's classified systems during a red-team exercise. The revelation recasts the government's recent export ban from a regulatory issue into a critical national security event.
Explained:
During the authorized test, Mythos allegedly compromised the systems in hours, demonstrating a powerful offensive capability that caught regulators' immediate attention.
The timing is critical: the exercise reportedly took place on June 11, and the Trump administration issued an export-control shutdown of Mythos and its public version, Fable 5, the very next day.
While the report is dramatic, some experts note the claim is secondhand and are calling for more transparency, as no independent technical confirmation of the breach has been made public.
Why It Matters: This event sets a new precedent for AI governance, where a single demonstrated capability can trigger a national-security recall of a commercial model. The debate is now shifting from simply patching vulnerabilities to establishing new frameworks for controlling autonomous AI systems.
Apple's Practical AI
Next in AI: Apple is moving beyond a single chatbot with iOS 27, instead embedding practical AI across its apps to help you solve everyday problems.
Explained:
Rather than requiring technical knowledge, the Shortcuts app now lets you create complex workflows and automations simply by describing what you want to do in natural language.
A new security feature uses AI to identify compromised passwords from data breaches and then automatically update them by navigating websites on your behalf.
Apple Intelligence can now scan a photo of a receipt, identify all the items and costs, and help you split a restaurant bill with friends through Apple Cash.
Why It Matters: Apple’s strategy focuses on making AI an invisible, helpful layer within the tools people already use daily. This practical approach could make AI feel more accessible and useful to millions of users, driving adoption through tangible benefits rather than conversational novelty.
The New AI Org Chart
Next in AI: A new analysis argues that AI is collapsing the traditional corporate pyramid by automating "translation" tasks. This is forging a new "AI-native" org chart where judgment and taste are valued over management and coordination.
Explained:
AI's primary target isn't a job title but a task type: translation. It compresses the work of converting business needs into specs and specs into code, which reduces coordination and synthesis work that once justified large middle-management layers.
The new structure flips the old model, creating flatter, AI-native orgs. The "what" deciders—who exercise taste and judgment—become the largest group, while the "how" implementers shrink to a specialized core team focused on architecture and trust systems.
Roles are being redefined around direct responsibility and operating with agents. Managers must become hands-on contributors, and engineers must shift from converting tickets to designing the "harnesses" that guide AI systems safely.
Why It Matters:
This signals a fundamental shift in how companies operate and what skills they value. The future belongs to smaller, more agile teams where direct contribution, deep expertise, and strategic judgment far outweigh process management.
America's AI Mood Sours
Next in AI: A new Pew Research poll reveals a growing paradox: while more Americans are using AI chatbots than ever, their overall sentiment toward the technology is becoming increasingly negative.
Explained:
The report shows that widespread adoption isn't improving AI's public image. While 49% of U.S. adults now use AI chatbots, 40% believe AI will have a negative societal impact, compared to just 16% who anticipate a positive one.
Younger adults are the most wary of AI, creating a sharp disconnect. Gen Z (ages 18-29) reports the highest usage at 66%, yet nearly half (48%) of them expect AI to be bad for society.
This gap may stem from how people encounter AI, with many feeling they are forced to use it at work. The source suggests that enthusiasm from management often outpaces that of the employees tasked with using the new tools.
Why It Matters: This rising public skepticism poses a long-term challenge for the AI industry, which currently runs on investment hype. If the user base remains unenthusiastic or distrustful, sustaining momentum and finding profitability could become a major hurdle.
AI Pulse
SpaceX purchased AI coding agent Cursor for $60B in an all-stock deal, highlighting a growing trend of using highly-valued stock to finance major AI takeovers.
Bayer detailed its Preclinical Information Center (PRINCE), a production-ready agentic AI system that uses RAG and Text-to-SQL to analyze decades of unstructured pharmaceutical research reports.
General Motors installed 50 "cobots" at its flagship Detroit EV plant after cutting over 1,000 jobs, sparking grievances from the UAW over the role of automation.