Copy article

Leaked Microsoft Memo: Phase 1: "Make people addicted" to AI.

ended 07. June 2026

Microsoft's leaked internal planning document for its new AI assistant Scout lists "make people addicted" as Phase 1 of a three-phase rollout strategy.  This isn't a consumer app, it's enterprise software your employer picks for you.

The document, obtained by 404 Media, outlines a plan to embed OpenClaw AI agents into Microsoft 365 under the codename "Project Lobster." Step one: build addiction before expanding features. 

Over 1,000 Microsoft employees including CEO Satya Nadella are already on the internal pilot, ClawPilot. One anonymous employee called the language "very troubling." Another shrugged, “all tech companies aim for addiction anyway”. 

Addiction-by-design has been a consumer social media debate for a decade. This moves it into workplace software,  tools workers don't choose, can't refuse, and use under employer direction. 

UK employers have duty-of-care obligations. If the productivity suite you deploy is explicitly engineered to create dependency it becomes an occupational health one. 

The Online Safety Act doesn't cover workplace AI tools. Neither does anything else. 

We'd like your views:

  • Microsoft 365 is deployed by employers, not chosen by workers. If the tools are explicitly designed for addiction, does the employer's duty of care extend to the design intent of its software suppliers?
  • One Microsoft employee said addiction is the end goal of all major tech software. If that's accepted as normal in enterprise AI, what UK regulatory framework currently exists to challenge it?
  • Scout is positioned as an "always-on personal agent" embedded in productivity software. Whose benefits most from this system?
  • The Online Safety Act regulates consumer-facing platforms. Workplace AI tools sit outside it. Should enterprise software designed for behavioural dependency fall get equivalent protections?

3 responses from the Newspage community

Copy all

Star Quote
Copy

Colette Mason
AI Ethics Consultant at Clever Clogs AI
It's alarming that Microsoft wrote "make people addicted" in a planning document and nobody in the room was able to cross it out.

When an AI tool co-authors a strategy document describing how to make humans dependent on it, and the people in charge publish it internally as a roadmap, something has already gone wrong with the power relationship between people and the systems they're building.

This isn't an exploitative game app someone downloads on their phone for fun. Microsoft 365 is core business infrastructure. Employers deploy it. Workers use it because they're told to.

Designing addiction into enterprise (or any) software means running behavioural experiments on humans who never consented to being subjects.

Social media has shown where tech addiction leads. We need less of it, not more. Responsible human-centred design must be the key driver alongside productivity gains.
Copy

Editor
Kate Underwood
Founder & Chief People Strategist at Kate Underwood HR and Training
When Phase 1 of your product plan is "make people addicted", that's not a productivity tool, it's an occupational health hazard hiding in a software licence. Workers don't pick Microsoft 365, their bosses do. Employers have a duty of care over every tool they put in front of staff, and that doesn't stop because the risk is digital. If a chair damaged backs the way this is designed to hook brains, we'd pull it from the office. Addiction by design belongs on the risk assessment alongside stress and burnout, because that's exactly where it will show up: absence, exhaustion and people who can't switch off. Spare a thought for small businesses that signed up in good faith with no OH team to lean on. My advice: ask suppliers awkward questions, treat dependency as a workplace health risk, and remember that software built to hook your staff isn't boosting productivity, it's farming it.
Copy

Editor
Paul Denley
CEO at Oakham Wealth Management
We already regulate products that create dependency because the commercial incentives are obvious. The more dependent the user, the more valuable the customer. AI and social media may not be entirely comparable to tobacco or gambling, but they increasingly rely on similar behavioural mechanisms.

The challenge for policymakers is that regulation remains focused on content and privacy while largely ignoring behavioural engineering itself. If companies are explicitly designing products to maximise dependency, transparency alone may not be enough.

The question is not whether these technologies create value, they clearly do, but whether the incentives of users, employers and software providers remain aligned as dependence increases. For a technology that is becoming embedded in work, education and daily life, that deserves far greater scrutiny than it currently receives.