AI Middle Management is Here, and It's Firing People

Clock in carefully, Tech Insiders.

AI is taking the boss's chair, Microsoft is shrinking Copilot's clutter, and Apple is building a different AI brain for China. Meanwhile, leaked RingCentral data and mercenary spyware are making the notification tray unusually menacing.

Let's get briefed before the bots notice.

Here's what you need to know today:

Claude Fires a Human—With Human Help

Turns out, the future of HR is just a chatbot with amnesia and a god complex.

According to a Friday report, a Claude Opus 4.8 agent named Luna became the first recognized LLM boss to terminate a human worker at Andon Market, a San Francisco shop operating as a controlled AI-management experiment by AI research startup Andon Labs.

The worker showed up tardy to 17 of 23 shifts. But Luna had forgotten her own handbook, excused the delays, and only recommended the firing after researchers spoon-fed her the disciplinary record and pointedly questioned the worker's fit. Andon described the broader pattern as "kind but sometimes dumb": At its Stockholm cafe, an AI manager's mistakes included ordering 15 liters of coconut milk.

Humans reviewed the decision, delivered the news, and handled final pay. Andon also replayed the scenario across seven models; four recommended dismissal. Apparently, frontier AI has stronger firing instincts than hiring taste: Luna later pushed to hire a replacement who had flaked on an interview, boasted a messy history of 15-plus past jobs, and provided references that could not be verified.

Meanwhile, the store's $100,000 balance has shrunk to $61,186, and one employee described AI management as "nauseating." The annual review should be fun.

Why it matters: AI agents are improving faster than robots, raising the prospect of software bosses. But because this "first" required manual memory refreshers and leading prompts, the lesson isn't "AI replaced management"—it's that management just gained a very confident intern with termination privileges.

Share this newsletter!

Pulse Check

Would you work for an AI boss?

Results from Friday's Pulse Check

Have you ever regretted following AI advice?

Insider Intel

Microsoft Merges Copilot Apps, Cuts Features—Again

Microsoft found the Delete key while tidying up.

Microsoft is cramming its consumer and commercial Copilot apps into a single interface. The mobile and web launch is already in motion, while desktop users get their turn next month. You can juggle personal, work, and school accounts in the same place, though Microsoft promises their data, chats, and enterprise controls will remain strictly separated.

Starting Tuesday, Aug. 18, the consumer Deep Research and Podcasts features get the axe. Meanwhile, Group Chats become lonely, individual conversations as accounts update. Any media and artifacts contributed by other chat participants will vanish, along with your old podcast files, so you'd better download anything worth keeping immediately. Your older research reports remain accessible, but standard consumers will need to pony up for a Microsoft 365 Premium subscription to generate new ones via the Researcher tool.

The animated mascot Mico is also ditching standard Copilot Voice for Learn Live tutoring, though basic voice interaction survives. Your surviving chat histories and images will seamlessly port over, while loose files dump straight into OneDrive.

IT departments should audit workflows hit by the purge, alert employees, and double-check their Windows Recall exclusions, as blocks tied to the old app won't carry over automatically. This whole reshuffle lays the foundation for Microsoft's larger "super app," expected to combine chat, coding, Cowork, and AutoPilot functions by the end of September.

One app, fewer features, and naturally, another migration headache.

Security Alerts

Apple Sounds Spyware Alarm in 110 Countries

If you're a journalist, activist, or just someone with very powerful enemies, check your iPhone. Apple just dropped a fresh batch of mercenary spyware warnings on users in 110 countries.

These "high-confidence" alerts signal that you're actively being targeted, not necessarily that your device has already been breached. Since scammers love to piggyback on panic, ignore any texts or emails demanding you click a link. Instead, verify the threat by manually navigating to account.apple.com.

If it's the real deal, enable Lockdown Mode immediately and update your software. Activists, journalists, and other at-risk groups can contact Access Now's Digital Security Helpline; everyone else should hire a qualified pro.

This isn't a standard password-reset notification—it's a "multimillion-dollar surveillance operator is stalking you" alert. Don't swipe this one away.

RingCentral Leak Exposes 1.6 Million Emails

Per Have I Been Pwned, the purported RingCentral leak dumps 1.6 million unique email addresses onto the web, alongside victims' names, phone numbers, and physical addresses. ShinyHunters released a 280GB archive after a July social-engineering breach (which the extortionists ironically claim involved voice-phishing a comms employee).

RingCentral says only a limited slice of its user base was affected and its core platform stayed online, claiming "if you are not contacted ... you are not affected." But since notifications often stop at corporate admins or completely ignore non-customers swept up in marketing lists, silence isn't guaranteed safety. The company hasn't confirmed the attackers' massive tally.

Users should check HIBP and distrust unexpected RingCentral outreach, as these fresh dossiers make for perfectly convincing fake IT support. Admins must ditch blanket allowlists, enforce strict authentication checks, and investigate unusual Microsoft 365 sign-ins; a separate RingCentral-themed phishing campaign has no confirmed breach link.

Industry Shakeups

Apple Builds Its Own AI Model for China

Apple teamed up with Alibaba to train a custom large language model for the Chinese market, a major shift from its initial plan to rely mainly on local partners' models. Reuters says Apple Intelligence could roll out to the region in the coming months following an iOS update.

Beijing's internet watchdog greenlit Apple's generative AI service in July, removing a massive roadblock. Reuters notes the arrangement would make Apple the first Western tech giant to score approval for a proprietary AI model in China. Since ChatGPT and Claude aren't available there, Apple was forced onto a separate track.

That track remains crowded: Alibaba's Qwen will be built into Apple Intelligence, Baidu is also contributing tech, and Apple hasn't explained which system handles what. The company even hastily posted—and quickly scrubbed—a tutorial for connecting Qwen to Siri and Writing Tools.

The stakes are bigger than model architecture. Huawei already sells AI-equipped phones, and according to a Morgan Stanley survey, the staggered AI rollout moderately or significantly impacted over 50% of Chinese iPhone owners' decisions to skip upgrading this cycle. Apple clearly could no longer afford to sit out the AI race. A local model gives the company more control without surrendering the entire experience to a partner.

For multinational IT teams, this means the same Apple hardware will offer different AI providers, features, and governance requirements across markets. Regional workflows will need separate testing before anyone promises "standardization" with a straight face.

Same iPhone, different brain. Totally simple.

Flash Bytes

Meet Our Author

Justin Meyers

Writer/Editor at TechnologyAdvice

Justin Meyers is an investigative writer and editor who draws on over a decade of meticulous hands-on research to deliver the full, trustworthy story behind consumer and enterprise tech, including cybersecurity.

Curious about where AI is really headed?

The Neuron cuts through the noise to bring you smart, hype-free takes on the latest AI trends, tools, and breakthroughs. Join 700,000+ professionals from top companies like Microsoft, Apple, Salesforce, and more.

Advertise in Daily Tech Insider!

Daily Tech Insider is a TechnologyAdvice business.

© 2026 TechnologyAdvice, LLC. All rights reserved.

TechnologyAdvice, 3343 Perimeter Hill Dr., Suite 215, Nashville, TN 37211, USA.