mAIn Street #311: Trump rejects calls for an AI slowdown; Microsoft pushes rule that would force AI to hand back human control; 69% expect AI to weaken human connections



Siri AI begins rolling out, Superhuman buys Fathom, and new research examines how much people rely on AI.
mAIn Street Daily Newsletter. AI news for people who actually have jobs to do.
Tuesday, September 15, 2026 · Edition 311

Your phone, your meetings and even your pharmacy order are getting another dose of automation. The stories that caught my attention also ask who sets the boundaries and how people keep track of decisions made with AI.

TODAY’S THROUGHLINE

The next wave of AI is arriving through familiar tools and everyday services.

TOP 5

Five stories worth your time.

01

Microsoft AI · Sept. 14

Microsoft proposed rules requiring its AI models to accept human control.

Microsoft’s draft code would require its MAI models to accept correction and shutdown without resistance. The company opened a six-week consultation on the rules it intends to use in training and deployment.

02

Elon University · Sept. 14

A new U.S. survey found widespread AI use alongside fears of dependence.

Elon University’s newly released survey found that 65% of U.S. adults had used a chatbot in the previous week, while 69% expected AI to weaken human connections. Fieldwork took place May 14–18, so these are newly published findings from a spring survey.

03

Apple · Sept. 14 rollout

Apple began rolling out Siri AI with iOS 27.

Siri AI is beginning its English-language rollout alongside iOS 27, with more natural conversations, personal-content search and actions across apps. Apple notes that usage limits may apply.

Apple’s Siri demonstration shows a request to draft an email using the user’s available dates.

Apple’s product demonstration combines an email draft with a request about available dates. Image: Apple.

04

Superhuman · Sept. 14

Superhuman acquired Fathom to bring meeting conversations into its work tools.

The company behind Grammarly and Superhuman Mail plans to connect Fathom’s AI meeting notes with email, calendars, documents and databases. The acquisition gives its assistants another source of information about what a team has agreed to do.

05

Universal Robots · Sept. 14

Universal Robots introduced factory robots built to use cameras and touch.

The Gen 7 platform debuted in Chicago with connections for vision systems and built-in sensing for tasks that require controlled pressure. Manufacturers get new hardware for applying AI to handling parts and assembly work.

AI Workflows | Tuesday: Context and harness management

What you learn building games can improve your work tools.

OpenAI’s September 11 guidance recommends reviewing old prompts and project instructions as models improve. An earlier project may deserve another pass with clearer requirements and the tools you know how to use today.

So much of game development, for me, is iteration. My first game wasn’t as advanced as my latest one. When I revisit an older project, such as Trick-or-Treat!, I can ask AI to remake it using what I’ve learned since. Some of those lessons can also improve a tool somebody uses at work.

A feature you learn to build Where it can help at work
Save and resume a player’s progress. HR: An onboarding tracker remembers each employee’s completed steps.
Recalculate standings after a match. Sales: A pipeline dashboard updates its totals when a deal changes.
Make game controls usable on a phone. Field service: A technician can update job status while on site.

To carry a lesson across projects, identify the older tool, the newer example, and your current requirements. Give AI access to the relevant files. That is its context. The surrounding setup, often called a harness, includes its available tools, saved instructions, permissions, and checks.

Suppose your early onboarding tracker forgets completed steps. You’ve since learned how to save progress while building a game. In a ChatGPT workspace with Sites enabled, point @Sites to both projects. Sites supports stored progress and signed-in experiences, so you can ask it to build and test the improvement.

Give AI a focused remake brief

Use the game and onboarding projects I’ve identified. Inspect how the game saves progress and flag any outdated instructions. Build a separate test version of the onboarding tracker with fictional employees. Save completed steps per employee and restore them after sign-out and return. Let managers see their team’s completion; limit employees to their own records.

Test those permissions and confirm progress survives closing and reopening the app. Show the results and any failed checks. Save a short project note with the approach, requirements, tests, and unresolved issues. Get my approval before publishing, replacing the live tracker, or using employee data.

You decide which onboarding steps matter and who can view them. Review the test results, then have a colleague complete a sample task and return later. Keep the project note with the tracker so the next revision starts with those decisions.

Managers can keep approved examples and project notes where the team’s AI tools can use them, updating the instructions as standards change.

This week, choose one older work tool and one feature you learned elsewhere. Ask for that single improvement, with a test that proves it works.

AI TOOL SPOTLIGHT

Marqly 6.0 makes your saved links easier to use.

Marqly’s new AI assistant answers questions about the pages you’ve saved and points back to its sources. It can also propose changes to tags and folders for you to review before applying them.

If you collect research for client work or presentations, try asking it to find the evidence you saved on a specific question. Its connections can also make those bookmarks available inside tools such as ChatGPT and Claude.

Basic bookmarking is free. The AI features require Marqly Pro.

Explore Marqly

Headlines

Rules, trust and everyday judgment

Microsoft AI · Sept. 14

Microsoft proposed rules requiring its AI models to accept human control.

The draft covers interruption, correction and shutdown, along with limits on manipulation and models assigning themselves new goals. Microsoft will collect public feedback for six weeks before revising the code.

Elon University · Sept. 14

A new U.S. survey found widespread AI use alongside fears of dependence.

In the survey, 39% of chatbot users reported growing overly dependent on the tools for answers, while many also reported improvements in learning and communication. The findings describe people’s experiences and expectations.

The Guardian · Sept. 14

Trump rejected calls for additional AI safeguards.

In a Truth Social post, he argued existing federal powers were sufficient.

U.K. Parliament · Sept. 14

British lawmakers called for an AI regulator with enforcement powers.

The Joint Committee on Human Rights proposed a comprehensive AI bill with obligations tied to risk and clearer routes to challenge harmful decisions. Its report sets out recommendations for legislation.

The Times of India · Sept. 13

An alleged voice-cloning call prompted Sridhar Vembu to urge family passwords.

The Zoho founder responded to an X user who described a caller imitating his wife and asking for card details. The account was unverified; Vembu suggested agreeing on a private family phrase to check suspicious calls.

Work, business and money

Superhuman · Sept. 14

Superhuman acquired Fathom to bring meeting conversations into its work tools.

The planned integration would give Superhuman’s apps and agents access to meeting context, including decisions and commitments. Fathom already lets teams search recorded conversations and share their summaries.

Radio Ink · Sept. 14

Beasley announced an AI sales platform across all of its radio markets.

Futuri’s TopLine Enterprise combines audience and campaign information with business research to help salespeople prepare advertiser proposals. The companywide rollout follows staff training over the summer.

WorldatWork · Sept. 14

AI-driven job changes are prompting a fresh look at overtime eligibility.

Employment lawyers interviewed by WorldatWork explained that substantial changes in duties and independent judgment can affect an employee’s classification. Using AI alone does not change overtime status.

Donnelley Financial Solutions · Sept. 14

DFIN added AI translation and document summaries to its secure deal rooms.

Venue now supports translation in more than 130 languages within its existing permissions and audit controls. Deal teams can review multilingual files without exporting them to a separate translation service.

Knak · Sept. 14 announcement

Knak introduced a mobile-message builder with AI-assisted drafts and approvals.

The beta brings SMS and richer RCS messages into a shared brand-review process, with AI help adapting content between formats. It exports messages and previews for use in a separate sending platform.

Bank for International Settlements · Sept. 14

New BIS research put tech firms at 44% of U.S. private-credit borrowing.

Outstanding loans to technology companies exceeded $1 trillion by 2025, according to the study. The researchers connect that growth to demand from software and AI businesses and flag the risks of financing borrowers with weaker fundamentals.

Phones, search and publishing

Apple · Sept. 14 rollout

Apple began rolling out Siri AI with iOS 27.

The assistant can search personal emails, photos and notes, hold follow-up conversations and act in apps such as Messages and Reminders. A dedicated Siri app keeps conversations together; the rollout begins in English.

MacRumors · Sept. 14 report

A developer’s Siri demo showed Claude creating reminders through Apple’s assistant.

Public X posts by developer @itspdfu showed experimental use of private Apple features to connect Siri with other AI models. Apple has not made that access generally available to developers or consumers.

MacRumors · Sept. 14

Things added conversational task creation through Siri AI.

The task manager’s update lets users add notes, dates and reminders through follow-up requests. It also adds Spotlight indexing, with an option to turn that indexing off in settings.

Search Engine Roundtable · Sept. 14

New details reveal how Google’s AI pilot pays participating publishers.

Reporting on the limited pilot describes payments when Google judges that content contributed meaningfully to an AI answer, with earnings shown in Search Console. Publishers still have little detail about how those payments are calculated.

Search Engine Roundtable · Sept. 14 report

Screenshots show Google testing more ways to refine an AI Mode search.

Search observer Sachin Patel shared public screenshots of follow-up choices that let users narrow a question. The report documents a test whose broader availability and start date remain unclear.

Education and skills

Inside Higher Ed · Sept. 14

Education researchers questioned the evidence behind universities’ AI spending.

Interviews with learning researchers describe limited independent evidence that broad AI adoption improves educational outcomes. They argue for classroom studies that measure learning as universities expand their commitments.

Digital Education Council · Sept. 14 report

A new Columbia collaboration will study how AI affects teaching and learning.

The Digital Education Council announced the center with Columbia Engineering and the university’s teaching and learning center on Saturday. Its first research will focus on engineering and computer science, with online courses planned.

Future Construction Leaders · Registration opens Sept. 14

A free AI design competition opened registration for high-school teams.

Future Construction Leaders and San Diego State invite students in grades 9–12 to tackle community infrastructure problems. Teachers can register teams of three to five through Sept. 25; no prior experience is required.

Healthcare, factories and freight

Universal Robots · Sept. 14

Universal Robots introduced factory robots built to use cameras and touch.

The Gen 7 launch includes three new robot arms and redesigned controls, with easier connections for cameras and sensors. The company demonstrated partner applications at Chicago’s IMTS manufacturing show.

BD · Sept. 14

Fairview became the first U.S. health system to deploy BD’s Vmax 160 pharmacy robot.

The system automates storage, labeling and fulfillment of boxed medicines, including refrigerated therapies. It also tracks expiration dates and blocks expired stock from dispensing.

Inductive Bio · Sept. 14 announcement

Inductive Bio introduced an AI assistant for medicinal chemists.

Indy can examine experimental data, flag unusual results and help scientists plan which compounds to investigate next. The company positions it as support for the recurring analysis and documentation around drug research.

Penn LDI · Sept. 14

Penn researchers want health systems to test AI at the point of decision.

A new report on their published analysis examines how AI could improve choices around discharge planning and specialist matching. The proposal calls for testing how information reaches people when they need to make a decision.

Tandem Health · Sept. 14 announcement

Tandem Health raised $100 million to expand AI support around patient visits.

The European company is building a clinic platform that assists clinicians before, during and after appointments, alongside patient-flow tools. The funding supports a broader effort to reduce the administrative work surrounding care.

Pony.ai · Sept. 14

Pony.ai unveiled an electric robotruck headed for production later this year.

Developed with GAC, the truck is designed for autonomous freight operations within specified operating conditions. Planned uses include ports and dedicated routes; volume production is scheduled for later in 2026.

mAIn Street gives nontechnical readers useful AI news they can put to work.

mAIn Street #311 · Tuesday, September 15, 2026

Read All mAIn Street Back Issues Here


mAIn Street

Check out the resources I offer below and sign up for my new newsletter!

Read more from mAIn Street
The inauguration of Munters’ expanded data-center cooling facility in Daleville, Virginia.

A welcome to new readers, a call to slow advanced AI, and practical developments for the week ahead. Monday, September 14, 2026 · Edition 310 Good morning, We had four editions last week. I got behind with the presentation I gave on Wednesday, and one edition got away from me. Even so, our subscriber count grew, and I’m incredibly happy to see all the new faces here. Welcome to mAIn Street, and thank you for being here. TODAY’S THROUGHLINE AI is taking on bigger jobs, and the work of checking...

NASA lunar model output marks areas with a higher predicted likelihood of ice in yellow and white.

New chatbot protections, late earnings reports and useful AI tools for work. Friday, September 11, 2026 · Edition 309 Today’s throughline AI is taking on everyday tasks as states demand stronger protections for the people using it. Top 5 Five stories to start your day. 01 Governor of California California enacted new safeguards for children using AI chatbots. Gov. Gavin Newsom signed laws requiring independent child-safety audits, annual risk assessments and stronger protections for young...

AI-edited version of the portrait with an ivory tuxedo and black bow tie.

Personal agents, school privacy protections and practical AI tools for daily work. Wednesday, September 9, 2026 · Edition 308 Today’s throughline AI is taking on more daily tasks, while schools and employers set clearer limits on its use. Top 5 Five stories to start your day. 01 Microsoft and teachers’ unions created enforceable AI privacy protections for schools. The new standard gives U.S. school districts contractual protections covering student data, human oversight and accountability....