Internal Microsoft documents have revealed that the company explicitly aimed to make users “addicted” to its AI assistant Copilot. This design philosophy has sparked widespread debate in the technology ethics community about whether AI products should balance user engagement with user wellbeing.

The Controversy

The documents show that Microsoft’s product teams are studying how to increase Copilot usage frequency through optimized interaction experiences, personalized recommendations, and instant feedback mechanisms. While similar strategies have long been criticized in social media platforms, explicitly targeting “addiction” as an AI product design goal is more contentious.

Critics argue that AI assistants, as productivity tools, should help users complete tasks more efficiently rather than encourage excessive engagement with AI interactions.

Broader Implications

This incident highlights a core tension in AI development: balancing user engagement with user wellbeing. The lessons from the social media era suggest that excessive pursuit of engagement can lead to filter bubbles, attention fragmentation, and mental health issues. AI assistants, being more personal and intelligent, may have even deeper “addiction” effects.

From a regulatory perspective, this event may accelerate legislation on AI product transparency requirements. The EU’s AI Act has already proposed transparency requirements for high-risk AI systems, and user engagement design may become a future regulatory focus.

Perspectives

Technology ethicists express concern. Several scholars note that AI assistant “addiction” design differs fundamentally from social media algorithm recommendations — AI assistants integrate more deeply into users’ daily workflows and habits, with more direct and lasting effects.

Microsoft has not officially responded. However, some industry analysts suggest the “addiction” language in internal documents may be a colloquial description of user engagement metrics rather than genuine intent to harm users.

Consumer protection organizations are calling for AI product design ethics standards, requiring tech companies to consider user wellbeing during the design phase rather than as an afterthought.

On April 29, 2026, Microsoft, Amazon, Alphabet, and Meta all reported their quarterly earnings simultaneously, presenting a complex picture of AI investment and returns.

Microsoft: Copilot Users Surge

Microsoft CEO Satya Nadella announced during the earnings call that Microsoft 365 Copilot now has 20 million paid enterprise seats. The number of companies paying for over 50,000 seats has quadrupled, with Bayer, Johnson & Johnson, Mercedes, and Roche each holding more than 90,000 seats.

Nadella emphasized that engagement levels now match Outlook: “Copilot queries per user were up nearly 20% quarter over quarter. Weekly engagement is now at the same level as Outlook. This is like a daily habit of intense usage.” Microsoft has also made Agent mode the default experience across Copilot, Word, Excel, and PowerPoint.

Morgan Stanley analyst Keith Weiss called the numbers “super impressive and way ahead of most people’s expectations.”

Amazon: AWS Strong Growth, But Free Cash Flow Under Pressure

Amazon Web Services (AWS) saw net sales increase 28% year-over-year to $37.6 billion, the fastest growth rate in 15 quarters. CEO Andy Jassy said, “We’ve never seen a technology grow as rapidly as AI.”

However, massive capital expenditures are pressuring free cash flow. Amazon’s trailing twelve-month free cash flow dropped to $1.2 billion, a 95% decline from $25.9 billion in the same period last year. Jassy argued this is a normal phase of high-growth cycles, noting that data center assets have useful lives exceeding 30 years.

Amazon’s overall revenue rose 17% to $181.5 billion.

Meta: AR/VR Burn Continues, Shares Slide

Meta’s Reality Labs division continues to post significant losses, and investor enthusiasm for AI spending is cooling. Following the earnings release, Meta shares slid, reflecting market concerns about the return timeline for Big Tech’s AI investments.

Overall, the Big Tech earnings results confirm strong demand for AI infrastructure investment, but also raise questions about whether the capital outlays will translate into sustainable revenue growth.

Source: TechCrunch - Microsoft, TechCrunch - Amazon, BBC

Microsoft-owned code hosting platform GitHub announced through its official blog that GitHub Copilot, its AI-powered programming assistant, will transition to a usage-based billing model starting June 1, 2026, replacing the previous quota-plus-overage system.

How the Billing Changes

Under the current model, Copilot users who exhaust their plan’s usage quota are automatically switched to a “Premium Requests” billing mechanism. The new usage-based approach introduces GitHub AI Credits — each plan tier will include a fixed quota of AI credits that are consumed as users interact with the service.

Notably, code completion and next-edit suggestion features will NOT consume AI credits. This means the core intelligent code autocompletion experience remains unaffected, while more advanced AI interactions — such as conversational coding assistance and complex code generation — will draw from the credit pool.

Pricing Remains Unchanged

GitHub confirmed that base pricing across all tiers will stay the same:

This pricing stability suggests GitHub is using usage-based billing to more equitably distribute computational resources, rather than as a pretext for price increases.

Industry Context

GitHub Copilot is one of the world’s most widely adopted AI programming assistants, with millions of developer users. The billing shift reflects a broader industry trend: as large language model inference costs remain high, AI service providers are transitioning from flat-rate subscriptions to more granular usage-based pricing.

Several AI service providers have already adopted similar credit-based or consumption-based pricing models. Analysts expect this move could trigger a ripple effect, accelerating the evolution of billing models across the AI developer tools industry.

Source: Solidot Report