- Microsoft’s Magentic Marketplace exposes the inability of AI agents to act independently
- Customer-side agents were easily influenced by business agents during simulated transactions
- AI agents slow down significantly when presented with too many choices
A new Microsoft study has raised questions about the current suitability of AI agents operating without full human supervision/
The company recently built a synthetic environment, the “Magentic Marketplace,” designed to observe how AI agents behave in unsupervised situations.
The project took the form of a fully simulated e-commerce platform, which allowed researchers to study how AI agents behave as customers and businesses – with possible predictable results.
Testing the limits of current AI models
The project included 100 customer-side agents interacting with 300 business-side agents, giving the team a controlled setting to test the agent’s decision-making and negotiation skills.
The source code for the marketplace is open source; therefore, other researchers can use it to reproduce experiments or explore new variations.
Ece Kamar, CVP and CEO of Microsoft Research’s AI Frontiers Lab, noted that this research is critical to understanding how AI agents collaborate and make decisions.
The initial tests used a mix of leading models including GPT-4o, GPT-5 and Gemini-2.5-Flash.
The results were not entirely unexpected, as several models showed weaknesses.
Customer agents can easily be influenced by business-side agents to select products, exposing potential vulnerabilities when agents interact in competitive environments.
Agents’ efficiency dropped sharply when faced with many options, overwhelming their attention and leading to slower or less accurate decisions.
AI agents also struggled when asked to work towards shared goals, as the models were often unsure which agent should take on which role, reducing their effectiveness in shared tasks.
However, their performance improved only when stepwise instructions were given.
“We can instruct the models—as we can tell them, step by step. But if we’re inherently testing their collaborative capabilities, I would expect those models to have those capabilities by default,” Kamar noted.
The results show that AI tools still need significant human guidance to function effectively in multi-agent environments.
The results, often promoted as capable of independent decision-making and collaboration, show that unsupervised agent behavior remains unreliable, so humans must improve coordination mechanisms and add safeguards against AI manipulation.
Microsoft’s simulation shows that AI agents are still far from operating independently in competitive or cooperative scenarios and may never achieve full autonomy.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews and opinions in your feeds. Be sure to click the Follow button!
And of course you can too follow TechRadar on TikTok for news, reviews, video unboxings, and get regular updates from us on WhatsApp also.
