Featured Image: yourNEWS Media Illustration
OpenAI is launching a dedicated election monitoring center ahead of the Nov. 3 midterms, bringing together hundreds of employees to identify potential misuse of ChatGPT following the discovery of Russian and Iranian influence operations using its technology.
By yourNEWS Media Newsroom
OpenAI is establishing an Election Operations Center to monitor potential interference in the Nov. 3 midterm elections, expanding its efforts to prevent artificial intelligence tools from being used for political deception, misinformation and coordinated influence campaigns.
The center will begin operating this week and remain active until vote counting is completed, according to a Wednesday report by Politico.
The initiative follows OpenAI’s disclosure earlier this month that influence operations originating in Russia and Iran had used ChatGPT to support deceptive political messaging, including fabricated documents, fictitious journalists and misleading research organizations.
Hundreds of OpenAI employees will participate in the election monitoring effort, which is designed to identify suspicious activity, investigate reports of unexpected AI behavior and respond to possible violations of the company’s policies.
Josh Lawson, OpenAI’s head of civic and strategic risk, will lead the operation.
“We’ve put considerable effort [into] ensuring we help users get reliable voting information, combat misuse of our tools and support election defenders,” Lawson told Politico.
“The Elections Operations Center builds on that yearslong work, bringing our teams together to respond quickly to emerging issues and enforce our policies as people head to the polls.”
The center will focus primarily on detecting misuse of OpenAI’s own products, including ChatGPT, rather than serving as a general authority over election information published elsewhere on the internet.
Its responsibilities include monitoring developing events, reviewing reports of suspicious activity and coordinating responses when users attempt to exploit AI systems for prohibited purposes.
The announcement follows an Oct. 8 OpenAI investigation detailing two influence operations that used ChatGPT alongside conventional methods to distribute political and geopolitical messaging.
The Russian operation reportedly created a purported research organization in Latin America, recruited individuals who appeared unaware of its underlying political purpose and used AI to prepare internal reports and fabricated materials.
Investigators found that the operation also generated false leaked documents and audio scripts, some of which circulated online.
The Iranian operation used seven fictitious journalist identities to submit articles to news organizations, primarily addressing the conflict involving the United States and Iran.
OpenAI identified nearly 100 articles published or republished under those identities across approximately a dozen online outlets.
The operators also used ChatGPT to prepare editorial pitches, refine political commentary and generate batches of social media posts.
However, the company reported that the social media activity generally attracted limited engagement.
OpenAI terminated accounts associated with both operations.
The company said the two cases illustrated how foreign influence actors were adapting established deception methods to take advantage of generative AI.
The findings did not establish that either operation had altered U.S. election results or successfully influenced a measurable number of American voters.
OpenAI has reported identifying 30 covert influence operations over approximately two and a half years, demonstrating the variety of attempts to exploit its technology for deceptive communications.
The new election center is intended to strengthen the company’s ability to respond to such activity during a period of heightened political attention.
In an Oct. 20 update outlining its election safeguards, OpenAI announced additional measures designed to help users obtain voting information and understand election results.
Through an integration with Democracy Works, ChatGPT provides information about voter registration, deadlines, polling locations and other election requirements, with links to official election authorities.
The company has also integrated election results from The Associated Press, allowing users to access live vote counts after polls close.
OpenAI emphasizes that state and local election officials remain the authoritative sources for election procedures and certified results.
Other safeguards address AI-generated images and deceptive political communications.
The company prohibits using its products to interfere with elections, discourage voting through deception or misrepresent the origin of coordinated political content.
It also restricts political advertising and the use of its systems to generate large-scale campaign persuasion materials.
OpenAI says it can respond to policy violations by restricting or terminating access to its services.
The company has additionally offered election officials and voting-system manufacturers access to AI tools intended to strengthen cybersecurity defenses.
The measures reflect growing concerns about the potential use of generative AI to create fabricated news reports, impersonate public officials or distribute misleading information about voting procedures.
Because AI systems can produce text, audio and images rapidly, they can reduce the time and resources required to develop deceptive material.
However, the discovery of an influence operation does not necessarily demonstrate that the operation reached a substantial audience or changed public behavior.
Researchers distinguish between attempts to manipulate public opinion, the actual distribution of deceptive content and evidence that such efforts influenced an election.
OpenAI’s election monitoring initiative is focused on identifying and disrupting attempts to misuse its technology before those activities spread further.
The operation will remain active beyond Election Day because vote tabulation, reporting and certification can continue after polls close.
With the Nov. 3 midterms approaching, the company’s Election Operations Center represents an effort to coordinate its existing safety, security and investigation teams during a period when misleading political information and attempts to exploit AI systems may receive increased attention.
Original article
Comments