Microsoft to restrict AI-generated images of candidates ahead of 2027 election — Official

Microsoft will restrict their generative AI tools from creating unrealistic images of political candidates ahead of Nigeria’s 2027 general election, a senior official has said.

Dave Liechtman, Microsoft’s Senior Director of Global Elections, disclosed this on Wednesday while speaking on a panel at the AI and Democracy Forum (AIDF) in Abuja.

The AIDF is a two-day event that brought together representatives of electoral commissions from West African countries, as well as security agencies, civil society organisations, the media, technology companies and students to discuss the role of AI in elections.

The forum, themed “Safeguarding the 2027 General Elections in the Age of Artificial Intelligence,” was organised by Yiaga Africa in partnership with the Independent National Electoral Commission (INEC), Microsoft, TikTok, the Centre for Journalism Innovation and Development (CJID), Dataphyte, the MacArthur Foundation, the International Foundation for Electoral Systems (IFES), the German Agency for International Cooperation (GIZ) and other partners.

Restricting AI images of candidates

Speaking during a panel session, titled “Will Algorithms Decide the 2027 Vote?”, Mr Liechtman said Microsoft would load INEC’s final list of candidates into its system to prevent generative AI from creating unrealistic images of the candidates.

“I offered to INEC at their offices this week that, – I think you just finalised the candidate list this weekend – we’re going to load that candidate list into our system to make sure that generative AI cannot be used to generate unrealistic images of candidates,” he said.

He said the initiative was part of the companies’ commitment to preventing the unauthorised use of candidates’ likenesses on their platforms.

Mr Liechtman said Microsoft and other AI-developing technology companies had committed to removing unauthorised synthetic media depicting candidates or election officials during elections.

He said the companies made the commitment in 2024, when technology firms signed an agreement known as the Tech Accord to Combat Deceptive Use of AI in Elections.

According to him, the agreement requires participating companies to remove content from their platforms that uses a candidate’s or election official’s likeness without authorisation.

“Google agreed to that, and all the platforms agreed to it. We have maintained that commitment for the last two and a half years, and we intend to continue maintaining that commitment,” he said.

Mr Liechtman acknowledged concerns about deepfakes but argued that their influence on elections might not be as extensive as some people fear.

He cited research from the United States suggesting that voters may distrust candidates more when they know the candidates have shared AI-generated synthetic media.

He said these findings should warn against circulating deceptive or misleading content.

“All candidates should take note of that and say, ‘Oh, when I’m sharing garbage out there, it actually turns people off to voting,’” he said.

AI and 2027 election

Mr Liechtman said AI, like other transformative technologies, could be used both constructively and harmfully, making it necessary to maximise its benefits while reducing the risks.

He said AI could have legitimate applications in elections and society but warned that safeguards were necessary to prevent its misuse.

READ ALSO: Iran threatens attack on facilities belonging to Google, Microsoft, other US/Israel companies

In his address, INEC Chairman Joash Amupitan said the electoral commission had established an AI unit to strengthen its digital infrastructure, including result management during elections.

He said AI-powered cyber-security systems could help detect denial-of-service attacks, brute-force attacks and other threats to electoral infrastructure.

In his welcome address, the Executive Director of Yiaga Africa, Samson Itodo, said rebuilding public confidence would be one of the major challenges ahead of the 2027 elections.

“Just last week, Yiaga Africa issued a national voting intention survey report, where it is obvious that Nigerians doubt the credibility of the process, and that their votes will not count in 2027,” Mr Itodo said.

“It is one thing to deploy technology; it is another thing for the personal integrity of those managing those technologies to deliver.”

Mr Itodo also raised concerns about the growing number of AI-powered platforms being developed to aggregate election results ahead of the 2027 elections.

He urged organisations deploying AI for results aggregation to ensure that their systems were accurate, transparent and auditable.

“For us, as an organisation, we’re very open and transparent about the methodologies that we use,” he said.