"Alarming critics, the acting director of the Cybersecurity and Infrastructure Security Agency (CISA), Madhu Gottumukkala, accidentally uploaded sensitive information to a public version of ChatGPT last summer, Politico reported. According to "four Department of Homeland Security officials with knowledge of the incident," Gottumukkala's uploads of sensitive CISA contracting documents triggered multiple internal cybersecurity warnings designed to "stop the theft or unintentional disclosure of government material from federal networks.""
"Gottumukkala's uploads happened soon after he joined the agency and sought special permission to use OpenAI's popular chatbot, which most DHS staffers are blocked from accessing, DHS confirmed to Ars. Instead, DHS staffers use approved AI-powered tools, like the agency's DHSChat, which "are configured to prevent queries or documents input into them from leaving federal networks," Politico reported. It remains unclear why Gottumukkala needed to use ChatGPT."
"That designation, a DHS document explained, is "used within DHS to identify unclassified information of a sensitive nature" that, if shared without authorization, "could adversely impact a person's privacy or welfare" or impede how federal and other programs "essential to the national interest" operate. There's now a concern that the sensitive information could be used to answer prompts from any of ChatGPT's 700 million active users."
Madhu Gottumukkala, acting director of CISA, accidentally uploaded sensitive CISA contracting documents to a public version of ChatGPT last summer. The uploads were marked "for official use only" and triggered internal DHS cybersecurity warnings intended to stop theft or unintentional disclosure of government material. Gottumukkala obtained special permission to use OpenAI's chatbot despite most DHS staff being blocked and the agency using DHSChat, which prevents queries from leaving federal networks. Some officials said he appeared to force CISA to grant access and then abused it. Officials worry the material could be used to answer prompts from any of ChatGPT's 700 million users, creating privacy and national-interest risks.
Read at Ars Technica
Unable to calculate read time
Collection
[
|
...
]