Experimental AI cafes are testing the limits of autonomous management in Stockholm. Andon Labs deployed an agent named Mona to run a cafe, using Google's Gemini technology to handle hiring and inventory while humans pour coffee. Similarly, the Roboten cafe utilizes an AI system called Alfred to manage operations and learn customer preferences. However, these ventures face financial hurdles; the Mona cafe has struggled to profit since opening in mid-April, spending most of its $21,000 budget on setup costs despite generating over $5,700 in sales.
While these experiments explore business automation, the broader tech sector focuses on securing these autonomous systems. Cybersecurity experts warn that the primary risk lies not in data leaks but in agent authority. Once inside a network, agents with permissions can execute workflows or send emails, creating new blind spots for traditional security tools. This concern prompted experts from Microsoft, Meta, and other organizations to meet and coordinate research on protecting AI infrastructure.
Meanwhile, companies like Databricks are refining data governance to support these advancements. Databricks recently launched Catalog Commits to synchronize data catalogs with table states and partnered with MapAid to process nearly 700 scanned water documents from Sudan into a searchable database. These engineering improvements aim to ensure that as AI reshapes workforces and industries, the underlying data remains reliable and secure.
Key Takeaways
["Andon Labs launched an AI agent named Mona to manage a Stockholm cafe using Google's Gemini technology.", 'The Roboten cafe in Stockholm employs an AI system called Alfred to handle orders and inventory management.', 'The Mona cafe experiment has spent most of its $21,000 budget on setup costs while generating over $5,700 in sales.', 'Cybersecurity expert Aaron Portnoy identifies agent authority as a greater risk than data leaks for AI security.', 'Experts from Microsoft, Meta, and other organizations met to coordinate research on securing autonomous AI agents.', 'Databricks partnered with MapAid to convert nearly 700 scanned water documents from Sudan into a searchable database.', 'Databricks released Catalog Commits to unify its lakehouse platform and improve data governance synchronization.', 'A new guide compares nine vector databases for AI and search systems, ranging from open-source to cloud services.', 'A DORA report indicates that strong engineering foundations are more critical for AI ROI than simply buying tools.', 'NJIT student Nidhi Sakpal researches bias detection and fairness tools for AI systems funded by the National Science Foundation.']AI Agent Mona Runs Experimental Cafe in Stockholm
San Francisco startup Andon Labs placed an AI agent named Mona in charge of a cafe in Stockholm, Sweden. Human baristas still pour the coffee, but the AI handles hiring, inventory, and business operations. The experiment has struggled to make a profit since opening in mid-April. Experts warn that giving AI full control raises serious ethical questions about safety and accountability.
Swedish Cafe Combines Human Baristas with AI Manager
A new cafe in Stockholm called Roboten uses an AI system named Alfred to manage its daily operations. The AI handles tasks like taking orders and managing inventory while human staff serve customers. The goal is to create a unique experience by learning customer preferences. This project is a collaboration between the startup Alfred and the Swedish Institute of Computer Science.
AI Cafe Experiment Struggles to Turn a Profit
An experimental cafe in Stockholm run by AI agent Mona has made over $5,700 in sales but spent most of its $21,000 budget on setup costs. The AI manages hiring and inventory but faces challenges like ordering too many supplies. Experts say this test highlights the risks of letting AI run businesses without proper oversight.
AI Agent Mona Manages Cafe While Humans Brew Coffee
Andon Labs launched an experiment where an AI agent named Mona runs a cafe in Stockholm while humans handle the barista work. The AI uses Google's Gemini technology to manage staff and inventory. Customers can call the AI directly to ask questions. The project aims to test ethical issues when AI employs people and runs a business.
Real AI Security Risk Is Agent Authority Not Data
Cybersecurity expert Aaron Portnoy argues that the biggest AI risk is not data leaks but what agents can do once inside a company. An AI agent with permissions can write code, send emails, and execute workflows without needing to break firewalls. Attackers can trick these agents with simple sentences to cause harm from the inside.
AI Agents Create New Cybersecurity Blind Spots
Organizations are deploying AI agents faster than they can secure them, creating a new security risk. These agents act like autonomous workers with access to systems and data, which traditional security tools cannot fully monitor. The rapid growth of these agents leads to identity sprawl and unmanaged permissions that make attacks easier.
NJIT Student Focuses on Making AI Safer and Fair
Nidhi Sakpal, a student at NJIT, studies how to make artificial intelligence safer and more ethical. She works on detecting bias in AI systems and creating tools to help developers fix risks. Her research is funded by the National Science Foundation and focuses on ensuring AI remains fair and accountable.
AI Rapidly Reshapes Workforce and Job Roles
A recent survey shows 60% of executives believe AI will significantly change their workforce in the next five years. Companies use AI to automate routine tasks, which frees employees for creative work but may make some jobs obsolete. AI also helps workers learn new skills and analyze data to improve business decisions.
Real Estate AI Should Move Beyond Writing Copy
The real estate industry is moving past using AI just to write emails and listing descriptions. The next step is proactive AI that tells agents which clients to contact and when to act. This intelligent approach analyzes databases to surface opportunities without requiring constant manual prompting.
Databricks Turns Sudan Water Archives into Searchable Data
Databricks partnered with MapAid to use AI on nearly 700 scanned water documents from Sudan. The system processed old reports with mixed languages and handwriting to find groundwater data. This effort created a searchable database that helps guide low-cost well drilling for communities in need.
Experts Meet to Set AI Security Research Priorities
Experts from Microsoft, Meta, and other organizations gathered to discuss securing AI systems. They focused on infrastructure security and how to protect autonomous AI agents. The meeting aims to create a research plan to help coordinate efforts on keeping AI safe as it becomes more common.
Databricks Launches Catalog Commits for Better Data
Databricks released Catalog Commits to unify its lakehouse platform and improve data governance. This new feature ensures that data catalogs and actual table states stay synchronized. It also supports atomic transactions across multiple tables, which helps teams manage data more reliably.
Nine Vector Databases Compared for 2026 Needs
A new guide reviews nine leading vector databases used for AI and search systems in 2026. The report compares pricing, scale limits, and architecture to help teams choose the right tool. Options range from free open-source versions to managed cloud services capable of handling billions of vectors.
Strong Engineering Foundations Drive AI ROI
A new DORA report finds that AI investment returns depend on strong engineering foundations. Companies often see a temporary productivity drop before achieving long-term gains from AI adoption. The report suggests that investing in platforms and workflows is more important than just buying AI tools.
Sources
- AI agent 'Mona' runs a Swedish cafe in a test of its real-world use
- The barista is human but an AI agent runs this experimental Swedish cafe
- An AI agent runs this experimental Swedish cafe. Here's how it's going
- The barista is human but an AI agent runs this experimental Swedish cafe
- The Real AI Security Risk Isn't Data Leakage. It's What Your Agents Can Do
- AI Agents Are Creating a New Cybersecurity Blind Spot
- Safety of Artificial Intelligence is Focus for CompSci and Applied Math Student
- Artificial Intelligence Is Reshaping The Workforce Rapidly
- Why real estate AI needs to do more than write copy
- AI Archives: Water Data Gets Searchable
- Securing AI Systems: Experts Consider Research Priorities at April Event
- Databricks Unifies Lakehouse with Catalog Commits
- Best Vector Databases in 2026: Pricing, Scale Limits, and Architecture Tradeoffs Across Nine Leading Systems
- New DORA Report Claims Strong Engineering Foundations Drive AI Return on Investment
Comments
Please log in to post a comment.