‘It can help you and it can hurt you’: The quandary of giving AI data and control

The way to get artificial intelligence to help you is to feed it more and more data and give it control, which on the flipside is also how it could ‘hurt’ you, said US computer scientist and podcaster, Lex Fridman. 

Fridman, expanding on how AI operates, said that AI agents get their power from the intelligence of the underlying model, how much access to data is given to the agents, and how much freedom and power is given. 

He said access to data and access to freedom are the biggest security problems. “And very soon, if not already, security will become THE bottleneck for effectiveness and usefulness of AI agents as a whole (1-3), since intelligence is still rapidly scaling and is no-longer an obvious bottleneck for many use-cases. The more data & control you give to the AI agent: (A) the more it can help you AND (B) the more it can hurt you,” he said.

Fridman said most tech-savvy folks are right now focusing on the useful part and not the security part. “I think solving the AI agent security problem is the big blocker for broad adoption. And of course, this is a specific near-term instance of the broader AI safety problem,” he said, adding that this is still the most exciting time to be alive for developers. 

Fridman’s remarks come when the world’s largest artificial intelligence summit is underway in Delhi. The India AI Impact Summit 2026 opened on Monday to packed halls and long queues, as tech moguls, industry leaders, policymakers, founders, and technologists at Bharat Mandapam in New Delhi. 

The summit is scheduled to be attended by Sundar Pichai of Alphabet to Sam Altman of OpenAI and Dario Amodei of Anthropic, apart from Prime Minister Narendra Modi, Brazil’s Luiz Inácio Lula da Silva and French President Emmanuel Macron. 

As many as 3,250 speakers across 500 sessions are lined up, while the expo features displays from Google to Nvidia, Amazon, Meta, OpenAI and Microsoft.