Aspen Digital

If you’re looking for more detail you can read our full notes below, or of course read the articles themselves. Feel free to send us suggestions at tom.latkowski@aspeninstitute.org.

Full Notes

‘DISINFORMATION on Steroids’: IS the US Prepared for AI’s INfluence on the Election? | US news | The Guardian

• Experts argue that US law and regulations are unprepared for the impact of AI on our elections
• Beyond just deepfakes, experts are concerned that the so-called “grandparent scam,” in which AI technology is used to mimic someone’s voice, could be targeted more widely
• Additionally concerning is the “liar’s dividend,” – the notion that because of fears of AI, people will stop believing true things they see or hear, allowing bad actors to avoid accountability

OpenAI, Meta, and other tech giants sign effort to fight AI election interference | Reuters

• In February, 20 major tech companies signed an accord at the Munich Security Conference, committing to create shared standards for watermarking AI-generated content and invest in public awareness campaigns

Opinion: AI doesn’t have all the answers — especially this election season | LA Times

• In an investigation led by Alondra Nelson and Julia Angwin, researchers found that when asked election-related questions, chatbots answered incorrectly 50% of the time
• Despite the fact that Google said its Gemini model would not respond to election-related questions, the investigators found that it did, and often gave false information

Generative AI RIsk Factors on 2024 elections | Munich Security Conference

• Aspen’s Vivian Schiller & Josh Lawson outline various risks at the intersection of AI and elections this cycle

Trump supporters target black voters with faked AI images | BBC

• Some supporters of former-President Donald Trump have shared AI-generated images of Trump with black voters in an effort to convince black voters to support Republicans
• Several false images have been widely shared on social media

The Shortlist: Seven Ways Platforms can prepare for the U.S. 2024 election | Protect Democracy

• The nonprofit group Protect Democracy announced its recommendations for mitigating the negative impacts of AI and technology on the 2024 elections
• (1) Platforms should invest in policy, legal, and trust/safety teams, and maintain these teams through inauguration day
• (2) Platforms should prominently offer accurate voting information
• (3) Usage rate limits should be enacted by platforms to reduce spam risks
• (4) Social media platforms should limit the distribution of new accounts and suspicious accounts in the runup to election day
• (5) Messaging platforms should ban “coordinated inauthentic behavior”
• (6) Generative AI platforms should disclose content authenticity
• (7) Generative AI platforms should prohibit their models from being used to disrupt elections, including through spreading falsehoods or intimidating officials

THE AI THAT COULD HEAL A DIVIDED INTERNET | TIME

• Jigsaw, a Google subsidiary, has used LLMs to create a set of tools that rank content based on characteristics like nuance, evidence-based reasoning, and compassion, rather than just on engagement.
• Jigsaw plans to make these new tools available for free to developers, in the hope that they’ll be popular with smaller websites
• Some people have raised concern that LLMs could bring biases to the ranking process, potentially further stigmatizing marginalized groups
• In a paper published by Jigsaw, users preferred comments sorted by their new classification algorithms then comments sorted by engagement

AI is changing how elections are fought, from deepfake endorsements to chatbot campaigners | ABC News

• Elections have already happened this year in Pakistan and Indonesia, and campaigning is ongoing in India, with numerous examples of AI usage
• In Indonesia, candidates are using a tool called “Pemilu.AI” as a data analytics tool to help them understand the voting public. Some candidates have used the tool as a partial replacement for campaign consultants
• Another Indonesian candidate — an ex-special forces commander who has faced questions of human rights abuses — is using AI-generated imagery to present himself as more friendly via a cartoon avatar
• In Pakistan, former Prime Minister Imran Khan is serving a prison sentence, but his party has released AI-generated speeches to mobilize his supporters
• In India, campaigners are using a deepfake of Muthuvel Karunanidhi, a politician who died in 2018. In the deepfake videos, Karunanidhi praises his son’s political career. Momium, the firm behind the deepfakes, has also used their technology to translate political speeches into multiple languages

Deepfake democracy: Behind the AI trickery shaping india’s 2024 election | Al Jazeera English

• On the morning of last years legislative elections, an Indian opposition party posted a deepfake video online showing a state political leader encouraging his supporters to vote for their party
• Many other Indian campaigns are deploying similar tactics, using deepfake videos to boost their chances of winning
• Some Indian politicians have used deepfake technology to generate videos of themselves speaking in multiple languages
• India has existing laws against defamation, but some experts have argued that they are insufficient to handle the onslaught of deepfakes

Countering Disinformation Effectively | Carnegie Endowment

• Researchers at the Carnegie Endowment put out a report on the most effective approaches to counter disinformation
• The authors noted that while there is not a “silver bullet” to solving the disinformation problem, policymakers can help address the problem by supporting local journalism, prioritizing fact-checking and media literacy, labeling false social media content, changing social media algorithms, and investing in counter-messaging strategies

Democratic operative admits to commissioning fake biden robocall that used AI | NBC News

• NBC news reported that Steve Kramer, a Democratic political operative, was the creator of the fake New Hampshire robocall which purported to be a recording of Joe Biden telling his supporters not to vote
• Kramer admitted that he was behind the call, stating that his goal was to encourage regulators to take action on the impact of AI in elections

NewsGuard launches suite of ai anti-misinfo tools | Semafor

• In March, the fact checking outlet NewsGuard launched a new effort to minimize the uses of AI to harm elections
• NewsGuard is investing in “fingerprinting” – identifying and labeling misinformation, information which can be fed into AI models to help them avoid sharing misinformation further
• NewsGuard’s 2024 Elections Misinformation Tracking Center will be a hub for this work

Election Disinformation takes a big leap with AI being used to deceive worldwide | AP News

• After AI deepfakes in Asian and European elections (including in Moldova, Slovakia, and Bangladesh), experts are increasingly concerned about the global impact AI-enabled disinformation
• FBI Director Christopher Wray echoed this sentiment, stating that generative AI would help “foreign adversaries to engage in malign influence”

Opinion: AI is turbocharging disinformation attacks on voters, especially in communities of color | LA Times

• AI-enabled disinformation could be a particularly big problem for voters of color
• Immigrants and communities of color are at particular risk for disinformation because they are more likely to face language barriers. With new AI tools, generating disinformation in foreign languages can be automated in ways that were previously impossible

Disinformation has a powerful impact on voting intentions | The Parliament Magazine

• Advocates are worried about the impact of AI-enabled disinformation in the runup to the June European Parliament elections
• Generative AI makes it easier and faster to generate disinformation, leaving democracies vulnerable
• Tech platforms, governments, and news organizations all have an important role to play in countering disinformation