google home ai cia response

Screengrab via Michael Hraba/YouTube

Google Home gets defensive when asked about the CIA

Well, that escalated quickly.


Phillip Tracy


A video on YouTube shows a Google Home providing an unnecessarily defensive answer to a man’s simple question about the CIA.

The video posted yesterday by Michael Hraba shows a man asking his voice-activated speaker if it knows what the agency is.

The inquiry appears to have struck a nerve with the assistant, which responded with a long-winded explanation of Google‘s stance on data privacy:

“No government entity, U.S. or otherwise, has direct access to our user’s information. Respect for the privacy and security of data you store underpins our approach to producing data in response to legal requests. You can learn more in our transparency report.”

It could be the AI still has those “Vault 7” documents WikiLeaks released earlier this month on its mind. Wikileaks said the CIA’s Engineering Development Group was overseeing “development, testing and operational support of all backdoors, exploits, malicious payloads, trojans, viruses and any other kind of malware used by the CIA in its covert operations worldwide.” The leaked files allegedly show the CIA’s Android and Chrome exploits, providing the name, descriptions, and types of devices the agency has affected.

Google responded to the documents, assuring its systems are not vulnerable to any exploits.

“We’ve reviewed the documents and we’re confident security updates and protections in both Chrome and Android already shield users from many of these alleged vulnerabilities,” said Heather Adkins, Google’s director of information security and privacy, according to Wired.

This isn’t the first time we’ve see an AI act out. Amazon’s Alexa couldn’t hold back when a young boy asked it to play “Digger Digger.” Its answer was a filth laden list of porn categories.

Update 5:10pm, March 23: A Google spokesperson provided the Daily Dot with a statement on why the speaker may have responded the way it did: “The Assistant responds to questions with a combination of algorithmically-generated answers and responses written by a team who helps craft the Assistant’s personality and answer specific questions aimed at the Assistant. In this case, we misunderstood the question. We’re looking into this response and strive to provide the most relevant answers to our users. We will continue to learn and improve the Assistant over time.”

Share this article

*First Published:

The Daily Dot