Welcome back to LLM Decode 👋
AI assistants are entering a new phase: more personal, more capable, and more deeply embedded in everyday technology. Apple is finally bringing its rebuilt Siri AI to users, while Microsoft is defining explicit rules for how increasingly capable AI should behave. Together, they highlight the next challenge: making AI more useful without giving up human control.
Here’s what matters today.
Apple’s Biggest Siri Upgrade Is Finally Arriving
Apple has begun rolling out its new software generation, with Siri AI as one of the headline upgrades. Apple originally unveiled the rebuilt assistant at WWDC26 as part of its next generation of Apple Intelligence, designed to make Siri significantly more capable and context-aware.
The biggest change is personal context. Siri AI can search across information in messages, emails and photos, understand what’s currently on a user’s screen, retrieve current information from the web, and perform more actions across apps. Apple has also introduced a dedicated Siri app where users can revisit previous conversations.
Apple is emphasizing privacy as part of the architecture, but availability depends on device, language and region. The company previously said the new Siri experience would initially require supported Apple Intelligence hardware, with some geographic restrictions.
Why it matters
AI is moving into the operating system. Users may increasingly interact with AI through their existing devices rather than opening a separate chatbot.
Personal context is the next battleground. Knowing your messages, files, apps and current screen can make an assistant substantially more useful.
Apps could become more action-oriented. The assistant layer can increasingly connect information with actions across software.
Distribution matters. Apple can introduce AI capabilities through an ecosystem of devices people already use every day.
Microsoft Is Writing a “Constitution” for Its AI
Microsoft has unveiled a draft AI code of conduct designed to act as a behavioral foundation or “constitution”, for future AI systems developed by the company. Microsoft AI CEO Mustafa Suleyman said the framework has been under development for roughly five to six months.
The rules are surprisingly explicit. Microsoft says its AI should accept human correction, remain understandable and never resist being interrupted or shut down. The company also states that its AI systems are not conscious and should not be treated as having legal personhood or rights.
The draft is now open for public feedback for six weeks. It arrives as leading AI companies and policymakers increasingly debate how to ensure that more autonomous and capable systems remain under meaningful human control.
Why it matters
AI behavior is becoming a design requirement. Companies aren't only training models to be smarter; they're defining boundaries for what those models should and shouldn't do.
Shutdownability matters more with agents. An assistant generating text poses different risks from an agent capable of independently taking actions.
Businesses can borrow the same principle. Define what an AI system may do, what requires approval and what it must never do before deploying it.
AI governance is moving into the product itself. Rules increasingly need to be implemented technically- not left as policy documents.
4 AI Tools & Community Workflows
Personal Context Workflow: Give your AI only the relevant emails, documents or project history for a task instead of repeatedly explaining everything from scratch. Apple’s Siri direction shows why context-aware AI is becoming increasingly valuable.
Screen-to-Action Workflow: When reviewing a webpage, report or dashboard, use AI to move directly from what you see → what it means → what action to take. This mirrors the shift toward assistants that understand what users are currently working on.
AI Permission Matrix: Divide agent actions into automatic / approval required / prohibited. For example, research can run automatically, sending an external email requires approval, and deleting business data is prohibited.
AI Kill-Switch Test: For every automated workflow, make sure a human can pause it, revoke access and review its actions. Microsoft's approach is a useful reminder that controllability should be designed before autonomy is increased.
Ready to level up your AI skills?
Explore Our Courses →That’s it for today.
The AI space doesn’t slow down - and neither should your thinking.
See you in the next drop.