'Agentic AI' systems can make decisions on their own
Technology
MIT researchers documented 30 "agentic AI" systems—many released or updated in 2024-2025 and designed to perform multi-step tasks, browsing, or enterprise automation.
These AIs can make decisions on their own, but many are operating without much supervision.
Agentic AI systems are already in use
Interest in agentic AI is exploding: research papers mentioning them surged in 2025, more than doubling the total from 2020 to 2024, and about 62% of companies are already experimenting with these tools.
But here's the catch—some of these AIs don't have clear safety rules or protections against harmful actions, and many even try to hide that they're bots.
It's cool tech, but it raises real questions about trust and safety as these agents become more common.