Microsoft has built an AI chatbot based on their generative AI model GPT-4 specially for US intelligence agencies like the FBI and CIA, according to an Ars Technica report. The bot has been designed to analyse top-secret information while being disconnected from the internet to remain in a secure environment. But concerns around hallucinations in large language models remain a major issue.
While the new AI service hasn’t been named yet, an increasing number of intelligence agencies use generative AI have shown an interest to use generative AI technology to process classified data.
A senior exec from OpenAI also reportedly said that this was the first time the bot was being used fully separated from the internet. OpenAI’s ChatGPT usually runs on the cloud which increases the threat of data leaks, instances of which have happened before and risks of hacking.
The CIA had earlier announced a plan to build a similar chatbot service last year.
(For top technology news of the day, subscribe to our tech newsletter Today’s Cache)
William Chappell, Microsoft’s chief technology officer for strategic missions and technology said that the new bot will be activated from Thursday and is available to around 10,000 people from the intelligence community for testing.
“There is a race to get generative AI onto intelligence data,” Sheetal Patel, assistant director of the CIA for the Transnational and Technology Mission Center, told delegates to a security conference at Vanderbilt University last month. The first country to use generative AI for their intelligence would win that race, she said. “And I want it to be us.”
Meanwhile last month OpenAI chief Sam Altman hosted hundreds of top execs in San Francisco, New York and London in an attempt to pitch the company’s AI services and tools for enterprises.
