How Social Media Algorithms Are Changing Modern Democracy for Better or Worse

Social media algorithms are changing modern democracy by deciding which political claims receive attention, which communities encounter one another, and how quickly false or emotional content spreads. Their effect is neither uniformly harmful nor automatically beneficial.

Recommendation systems can help people find news and civic information, but they can also reward material that produces strong reactions, narrow a person’s information diet, and make public debate harder to share.

The most accurate answer is that algorithms do not control voters by themselves, yet they can shape the conditions in which people form opinions, judge evidence, and participate in elections.

What social media algorithms mean for democracy

• Algorithms rank content rather than showing a neutral timeline. They commonly use signals such as previous activity, predicted engagement, relationships, content characteristics, and user feedback. The exact systems vary by platform and change over time.
• The strongest evidence does not support one simple claim that algorithms alone cause polarization. Political identity, partisan media, social networks, offline events, and platform design all matter.
• Recommendation systems affect visibility. A claim does not need to persuade everyone to matter politically. It only needs to receive enough distribution to reach journalists, activists, influencers, and voters.
• Engagement is not the same as public importance. A post can perform well because it provokes anger, fear, or ridicule rather than because it is accurate.
• Social media can improve civic access. It lets candidates, public agencies, journalists, and ordinary citizens publish directly to large audiences. Pew Research Center reports that social media remains an important source of news for many adults, although usage differs by country and platform. (Pew Research Center)
• The best response is not simply to remove all recommendation. Better options include meaningful user controls, independent audits, political ad transparency, stronger media literacy, access for qualified researchers, and clear appeals for moderation decisions.

What is a social media algorithm?

A social media algorithm is a set of computational rules used to select, rank, recommend, remove, or otherwise organize content. On a typical platform, the system may choose a small selection from a much larger pool of posts, videos, comments, or accounts.

The term “algorithm” can sound more mysterious than it is. In practice, a recommendation system often tries to predict what you may watch, read, like, share, comment on, or continue viewing. It can also account for safety rules, legal requirements, commercial goals, and the content’s relationship to your interests.

Platforms do not all use the same signals. Meta says its systems use various predictions and signals to rank content in services such as Facebook and Instagram, while TikTok describes a recommendation process that considers user interactions, video information, and device or account settings. (Meta Transparency Center, TikTok Newsroom) These explanations describe broad categories rather than publishing every operational detail.

A simplified example looks like this:

1. You watch several clips about a national election.
2. The platform records signals connected with that activity.
3. It predicts which related clips may hold your attention.
4. It places some of those clips in your feed.
5. Your next actions generate new signals.
6. The system updates its predictions.

This creates a feedback loop. Your choices influence what is shown to you, but what is shown also influences the choices you can make next.

That distinction matters. A person is not merely receiving a fixed media diet. The platform is continually arranging the menu.

Why do recommendation algorithms matter politically?

Recommendation systems matter because attention is a political resource. The content that reaches a person first can influence which issues appear important, which explanations seem plausible, and which groups appear trustworthy or threatening.

A recommendation system does not need to persuade a citizen directly to have political effects. It may affect democracy by:

• making one issue more visible than another;
• introducing a person to a political creator;
• repeatedly presenting the same interpretation of an event;
• moving a claim from a small online group into mainstream discussion;
• making a correction less visible than the original false claim;
• connecting people to news during an emergency or election;
• allowing campaigns to target narrow audiences with tailored messages.

The effect can be positive. A local resident may learn about a public meeting, a voter may find registration information, or a minority community may gain a channel to organize when traditional media ignore it. The effect can also be negative when the system repeatedly rewards inflammatory content and reduces contact with reliable reporting.

The political question is therefore not simply, “Are algorithms biased?” A better set of questions is:

• What does the system optimize?
• Which users receive which content?
• Can people understand or change the ranking?
• What happens to low-engagement but important information?
• Who can independently test the platform’s claims?
• How quickly can false or harmful content spread?
• Does the system treat political content differently from entertainment?

These questions move the discussion away from a vague fear of technology and toward measurable design choices.

Do social media algorithms cause political polarization?

The short answer is: they can contribute to polarization, but available evidence does not show that recommendation algorithms are the sole or universal cause.

Political polarization has several forms. Ideological polarization means people hold increasingly different policy positions. Affective polarization means people develop stronger dislike or distrust toward supporters of another political group. Social media can influence both, but the pathways are not identical.

One pathway is selective exposure. People may prefer content that confirms their existing views, and platforms may then recommend more of what appears to match those preferences. This can produce a narrower information diet. Yet the direction of causation is difficult to establish: users’ political identities may shape the recommendations they receive just as recommendations shape later behavior.

A major study of Facebook’s U.S. users during the 2020 election found that changing what people saw in their feed could alter exposure to politically aligned content, but the research did not find that switching the ranking system substantially changed key political attitudes during the study period.

The findings were published in Nature in 2023 by researchers including Talia Stroud and colleagues. (Nature, “How do social media feed algorithms affect attitudes and behavior in the 2020 U.S. presidential election?”)

Another Nature study from the same research project found that changing Facebook’s feed ranking substantially changed the content users saw, including their exposure to politically aligned content, but did not produce a measurable change in political attitudes over the study period. (Nature, “Like-minded sources on Facebook are prevalent but not polarizing,” 2023)

Those findings should not be overextended. They do not prove that all platforms, countries, election periods, or recommendation systems have the same effect. They also examined a defined period and specific measures. Still, they challenge the simple claim that changing a feed automatically changes voters’ political beliefs.

The evidence points to a more complicated pattern:

• algorithms influence what people encounter;
• users bring existing identities and preferences;
• political elites and media organizations create much of the content;
• social networks determine which posts receive early attention;
• offline events can overwhelm online effects;
• repeated exposure may affect salience even when it does not change a person’s core ideology.

In other words, recommendation systems may be better at shaping attention and perceived importance than at converting a committed voter from one political side to another.

How does algorithmic amplification differ from polarization?

Algorithmic amplification means a system gives content more distribution than it would receive through direct follower relationships or chronological posting. Polarization is a change in political attitudes or group relations. The two can occur together, but they are not the same.

For example, a platform may amplify a hostile post because many users comment on it. That may increase the post’s reach without changing everyone’s political position. However, repeated exposure to hostile portrayals of an opposing group could strengthen negative feelings, particularly among people who already identify strongly with one side.

Researchers must therefore distinguish among:

• exposure: what a person sees;
• engagement: what a person clicks, watches, likes, or shares;
• belief: what a person accepts as true;
• emotion: how a person feels about a political group;
• behavior: whether a person votes, donates, protests, or attacks others.

A platform may strongly affect one measure and barely affect another. That is why broad claims about “social media changing democracy” need precise evidence.

Do algorithms create filter bubbles?

A filter bubble is a personalized information environment in which a person receives less exposure to viewpoints or facts that challenge existing beliefs. The idea is plausible, but it is often presented too simply.

People build information environments through more than automated recommendations. They choose whom to follow, which groups to join, which news sources to trust, and which posts to ignore. Friends, family, political parties, influencers, search engines, and news organizations all affect what they encounter.

A user may also receive politically diverse content without accepting it. Seeing another viewpoint does not mean giving it equal credibility. Conversely, following multiple sources does not guarantee genuine diversity if those sources repeat the same evidence or assumptions.

The strongest answer is that platforms can create partial and uneven filter bubbles, but no single platform produces the same bubble for every user. The result depends on:

• the platform’s ranking system;
• the person’s past behavior;
• the accounts they follow;
• the political context;
• language and location;
• the availability of independent news;
• whether the platform recommends content outside a user’s network.

A 2015 Facebook study by Eytan Bakshy, Solomon Messing, and Lada Adamic found that users’ own choices played a larger role in limiting exposure to ideologically different content than the algorithmic ranking system did, although the ranking system also contributed. (Science, “Exposure to ideologically diverse news and opinion on Facebook,” 2015)

That result does not settle the issue today. Platforms have changed, political behavior has changed, and short-video recommendation systems operate differently from the Facebook feed studied at the time. Still, it offers an important lesson: users and algorithms interact. Blaming only the machine hides the role of human choice, political identity, and social networks.

Why does engagement often favor emotional political content?

Engagement is easy for a platform to measure. Trustworthiness, civic importance, and long-term social effects are harder to measure. This difference can create a problem when systems rely heavily on predicted interaction.

Emotion is not always harmful. A powerful account of injustice may help a neglected issue receive attention. A warning about a dangerous event may need rapid distribution. Satire and humor can help people participate in politics.

The risk appears when anger, fear, humiliation, or shock consistently produces more measurable activity than careful explanation. A misleading post can attract thousands of comments because people are correcting it, mocking it, or fighting over it. A calm fact-check may receive less visible activity even though it is more useful.

A 2021 study in Science found that political content on Facebook and Instagram was more likely to receive high engagement when it used language associated with moral and emotional expression. The study analyzed millions of posts from members of Congress and other political actors. (Science, “Moralization and polarization in online political discourse,” 2021)

The precise relationship between emotional language and algorithmic ranking varies by platform. Nevertheless, the measurement problem remains: a system that treats attention as a proxy for importance may distribute content that keeps people reacting rather than content that helps them understand.

This can affect democratic discussion in three ways.

1. It changes the speed of political claims

A claim can reach a large audience before journalists, researchers, or public agencies verify it. Corrections often arrive later, and later information may not receive the same attention.

2. It changes the style of political communication

Candidates and activists may learn that a short, dramatic accusation performs better than a careful policy explanation. Over time, this can make public debate more theatrical and less informative.

3. It changes the cost of participation

People who want to speak responsibly may face pressure to use stronger language simply to remain visible. The result is not necessarily that every citizen becomes extreme. It may mean that moderate or careful views become less prominent.

How do algorithms affect misinformation and disinformation?

Misinformation is false or inaccurate information shared without a proven intent to deceive. Disinformation is false or misleading information distributed deliberately to deceive or manipulate. The distinction concerns intent, which is often difficult to verify from a post alone.

Algorithms can affect false claims by influencing their reach, repetition, and audience. They do not create every false claim. Political actors, state-linked operations, opportunistic influencers, ordinary users, and coordinated groups may produce or distribute misleading material.

Research on the spread of false news on Twitter found that false news traveled farther, faster, deeper, and more broadly than true news in the dataset studied. The authors, Soroush Vosoughi, Deb Roy, and Sinan Aral, published the study in Science in 2018. They reported that false political news showed a particularly strong difference in diffusion. (Science, “The spread of true and false news online,” 2018)

That study examined Twitter activity from 2006 to 2017 and did not prove that algorithms alone caused the results. Human users shared the material, and platform ranking, network structure, novelty, and emotional responses may all have mattered. The findings nevertheless show why false political claims can become widely visible before verification catches up.

The World Health Organization uses the term “infodemic” to describe an overabundance of information, including false or misleading material, during a disease outbreak. Its guidance explains that an infodemic can make it difficult for people to find trustworthy sources and reliable guidance. (World Health Organization, “Infodemic”)

The same basic problem can occur during elections. Voters may face false information about:

• voting dates and registration;
• polling locations;
• eligibility rules;
• candidates’ statements;
• election results;
• public safety;
• fabricated endorsements;
• manipulated audio, video, or images.

The danger is not only that people believe one false story. A sustained flood of conflicting claims can create confusion and cynicism. A person who concludes that “nothing can be trusted” may withdraw from informed participation even if they do not accept a particular falsehood.

Can artificial intelligence make algorithmic political manipulation worse?

AI-generated text, audio, images, and video can reduce the cost of producing convincing political material. This does not mean every synthetic media item will influence voters. Its effect depends on timing, distribution, plausibility, existing beliefs, and whether the content is identified quickly.

The danger is highest when synthetic content appears close to an election, targets a narrow audience, or exploits a crisis. A fake recording may not need to convince the entire public. It may only need to confuse a small group, damage trust in a candidate, or give supporters a reason to reject legitimate evidence.

The Federal Communications Commission ruled in 2024 that AI-generated voices in robocalls fall under the Telephone Consumer Protection Act, following a fake voice message that imitated President Joe Biden during the New Hampshire primary season. (Federal Communications Commission, “FCC Makes AI-Generated Voices in Robocalls Illegal,” 2024)

The incident shows how synthetic media can combine with familiar political tactics. The technology may be new, but impersonation, voter suppression, deception, and targeted persuasion have older histories.

For readers, the practical rule is simple: treat emotionally urgent audio or video as unverified until you can confirm it through an independent, reputable source. A label from a platform can help, but it should not replace outside verification.

How do recommendation systems change what people think is important?

Public opinion is shaped not only by what people believe but also by what they believe matters. This is where algorithmic ranking can influence politics without changing a person’s ideology.

Suppose a platform repeatedly places crime, immigration, corruption, or inflation near the top of a user’s feed. The user may conclude that the issue is widespread, urgent, or ignored by authorities. That perception can affect voting priorities even if the person’s underlying policy views remain stable.

This is related to agenda-setting, the idea that media attention can influence which issues people perceive as important. Recommendation systems add a personalized layer: two citizens may receive very different signals about what deserves attention on the same day.

That can weaken a shared public agenda. Democratic disagreement is normal, but citizens need some common reference points to debate laws, evaluate officials, and respond to public emergencies. Personalization can make public attention more fragmented.

Still, fragmentation is not always a loss. A national news agenda may overlook local issues, minority concerns, or specialized policy questions. Recommendation systems can help those subjects find an audience. The challenge is deciding whether personalization expands civic knowledge or traps people in narrow streams of emotionally selected content.

A useful test is to ask:

• Can users tell why they are seeing a post?
• Can they switch to a chronological or non-personalized feed?
• Can they intentionally request local, international, or opposing viewpoints?
• Does the platform recommend primary documents and reliable reporting?
• Does it reduce the reach of content that violates its rules?
• Are users given context when a claim concerns an election or public emergency?

These design choices matter more than the word “algorithm” by itself.

What role do algorithms play in elections?

Elections place unusual pressure on information systems. The time available to verify claims is short, public attention is high, and a small change in turnout can matter.

Recommendation systems influence elections through several channels:

Candidate visibility

A candidate with a strong presence on a platform may reach voters without relying on traditional news coverage. This can lower the cost of political communication and give lesser-known candidates a route to public attention.

However, visibility is not evenly distributed. Platform rules, creator networks, advertising budgets, language, geography, and existing popularity can all affect reach.

Political advertising

Targeted advertising allows campaigns and outside groups to show different messages to selected audiences. This can make public scrutiny harder because the wider public may not see every claim.

The European Union’s Digital Services Act requires very large online platforms and search engines to provide information about certain advertising systems and risks, while the European Commission has also developed transparency rules for political advertising. (European Commission, Digital Services Act, European Commission, political advertising regulation)

Rules differ by country, and enforcement is still developing. Therefore, readers should not assume that a label such as “sponsored” reveals the full targeting logic or funding chain.

Turnout information

Accurate reminders about registration, deadlines, and polling places can support participation. False information about these matters can discourage or misdirect voters.

The U.S. Cybersecurity and Infrastructure Security Agency states that election infrastructure includes systems connected to voter registration, voting, vote counting, and election results, and it provides public guidance on election security. (CISA, Election Security)

For election information, I recommend checking the official election authority in your jurisdiction rather than relying on a viral post, even when that post appears to come from a familiar account.

Crisis communication

During contested elections, official results may develop over hours or days. Recommendation systems can circulate preliminary claims, rumors, and legitimate updates at the same time. A responsible platform needs clear labels, reliable links, and rapid responses to false voting instructions without suppressing legitimate reporting.

Can social media strengthen democracy?

Yes. Any serious discussion should include the benefits.

Social platforms can:

• give citizens direct access to elected officials and public agencies;
• help local groups organize around public issues;
• provide rapid updates during emergencies;
• allow journalists to distribute reporting beyond a single outlet;
• help marginalized communities document events and find one another;
• create public records of speeches, promises, and official conduct;
• let candidates communicate without traditional gatekeepers;
• support civic education through explainers, debates, and live events.

The Reuters Institute Digital News Report has documented the growing role of social video, online platforms, and creators in news consumption across many countries. Its findings also show that audiences often encounter news incidentally through platforms rather than deliberately visiting news websites. (Reuters Institute, Digital News Report 2024)

That shift can widen access. A person who never buys a newspaper may still see a public health notice, an interview with a candidate, or an explanation of a proposed law.

The benefit depends on information quality and user control. Direct access can remove editorial gatekeepers, but it can also remove useful verification. The answer is not to return to a single information authority. It is to make the provenance, evidence, funding, and correction history of political content easier to inspect.

Why are “neutral algorithms” not a complete solution?

A platform might say it wants to rank content neutrally. Yet every ranking system makes choices.

Chronological ordering favors recent posts. Engagement ranking favors activity. Personalized ranking favors predicted relevance. Editorial curation favors human judgment. Safety ranking favors risk reduction. Each approach gives some values more weight than others.

Even a system that does not use political ideology directly can produce political effects. For example:

• prioritizing video may favor emotional performance over detailed text;
• prioritizing recent activity may reward coordinated posting;
• prioritizing comments may reward conflict;
• prioritizing popular accounts may concentrate political visibility;
• reducing links may weaken the distribution of independent journalism;
• downranking uncertain claims may also reduce legitimate breaking news.

This does not mean platforms should give every claim equal reach. Accuracy, safety, privacy, and free expression can conflict. The important point is that ranking is a form of governance. Platforms make practical decisions about public visibility, even when they describe themselves mainly as technology companies.

The Stanford Internet Observatory’s work on platform governance has emphasized the importance of access to data and independent research when assessing these systems. (Stanford Internet Observatory, research and publications)

What does platform transparency need to include?

A transparency report that lists the number of removed posts is useful, but it does not answer every democratic question.

A more complete system would explain:

1. What signals influence recommendations?
   Users need understandable categories, not only technical language.
2. How do political and public-interest posts receive distribution?
   A platform should clarify whether it treats news, campaign content, satire, and civic information differently.
3. What happens when a user changes settings?
   If a person chooses chronological order or a non-personalized feed, the platform should explain what changes.
4. How are errors corrected?
   Users need accessible appeals, especially when moderation affects journalism, political participation, or public safety.
5. Can independent researchers inspect outcomes?
   Aggregate access can reveal whether certain groups, languages, or topics receive unequal treatment.
6. How are advertisers identified?
   Political ad archives should show sponsor information, spending, audience criteria, dates, and the creative material.
7. How are recommendation risks tested before deployment?
   Companies should assess foreseeable effects on minors, elections, public health, and targeted harassment.

Transparency is not the same as publishing source code. Code can be difficult to interpret, can expose security risks, and does not necessarily reveal how the system behaves in real use. Outcome data, controlled testing, documentation, and independent oversight may provide more practical information.

What can you do to protect your information diet?

Individual action cannot fix platform design, but it can reduce the chance that a recommendation loop becomes your only view of public life.

1. Separate discovery from verification

Treat a post as a lead, not as proof. Open the original source. Check the date, location, author, and supporting evidence. For an election claim, consult the relevant election authority.

2. Read beyond your preferred feed

You do not need to give equal time to false claims. However, you can compare reputable sources that use different editorial perspectives. Look for agreement about basic facts before debating interpretation.

3. Check the source, not only the screenshot

Screenshots can omit context and may be edited. Find the full article, video, transcript, court filing, government document, or research paper.

4. Pause before sharing outrage

Ask what the post wants you to do. Does it provide evidence, or does it mainly provoke immediate anger? Emotional force is not proof of deception, but it is a reason to slow down.

5. Use platform controls

Where available, adjust topics, mute accounts, unfollow without blocking, reset recommendations, choose a chronological feed, or mark content as irrelevant. These controls do not remove every influence, but they can change the signals the system receives.

6. Keep a direct list of reliable sources

Follow public agencies, local journalists, election offices, research organizations, and news outlets directly. This reduces dependence on whatever the recommendation system happens to select.

7. Do not quote a false claim without context

Repeating false information can increase its visibility even when your intention is to criticize it. Link to a correction and explain the evidence rather than reproducing a sensational headline when possible.

8. Be careful with AI-generated media

Look for independent confirmation, unusual editing, missing context, and claims that appear only in one place. Reverse image searches can help, but they are not conclusive. When the claim is important, wait for confirmation from multiple reliable sources.

What should policymakers and platforms change?

A democratic information system should protect expression while reducing preventable manipulation. No single intervention will solve the problem.

Give users meaningful choice

Users should be able to select feeds based on clear options, such as chronological order, accounts they follow, local public information, or personalized recommendations. The choice should be understandable and easy to change.

Require political advertising records

People should be able to see who paid for an ad, the dates it ran, the audience categories used, and the versions shown to different groups. Researchers should be able to study patterns without exposing private personal data.

Support independent research

Companies should provide privacy-protective access to data for qualified researchers. Without outside testing, the public must rely mainly on company statements and limited academic studies.

Protect researchers and journalists

Researchers who document platform harms should not face unclear enforcement or arbitrary account removal. Platforms still need rules against unauthorized access and personal-data abuse, but those rules should not block legitimate public-interest investigation.

Build better election information systems

Platforms can direct users to official election information, label altered media, limit false voting instructions, and publish rapid reports during election periods. These measures should be specific and evidence-based rather than broad bans on political disagreement.

Measure more than clicks

If a platform rewards watch time and comments, it should also test whether users can identify reliable information, encounter important civic material, and understand why they are seeing a post. Public accountability should include meaningful outcomes, not only usage numbers.

Preserve room for human judgment

Automated systems can process huge volumes of material, but they often struggle with satire, local context, language variation, and rapidly changing events. Human review remains necessary for high-impact decisions, particularly when content affects elections, public health, or journalism.

What does the research still not tell us?

Several questions remain difficult to answer with certainty.

First, researchers often cannot observe the complete recommendation system. Platforms may disclose broad principles but not all ranking signals, experiments, or enforcement changes.

Second, people who use different platforms may already differ in age, political identity, education, location, and media habits. Comparing them does not automatically prove that the platform caused their political views.

Third, engagement data can be misleading. A comment may represent agreement, anger, ridicule, or an attempt to correct a false claim. A view does not prove belief.

Fourth, effects may differ by country. Language, party systems, media institutions, election rules, internet access, and trust in government all shape the result.

Fifth, short-term exposure may not produce an immediate change in a survey response. A platform might influence which issue a person discusses, whether they contact another voter, or how much they trust institutions, without changing a standard ideology measure.

Accordingly, the most responsible conclusion is not that algorithms are harmless or all-powerful. It is that they are part of a larger political communication system, and their effects depend on design, users, institutions, and events.

A practical test for evaluating a political post

When I encounter a political claim online, I use this five-step check:

1. What exactly is being claimed?
   Rewrite it in one sentence without the emotional language.
2. What is the original source?
   Identify the person, organization, document, or data behind it.
3. When was it published?
   An old image or statistic can become misleading when presented as current.
4. What evidence supports it?
   Look for primary documents, named research, official records, or direct quotations.
5. What do independent sources say?
   Compare at least two reputable sources that are not simply copying one another.

For election procedures, I give greater weight to official election authorities. For scientific claims, I look for the original study and reputable scientific institutions. For breaking news, I check whether multiple established news organizations have independently confirmed the report.

This process takes longer than sharing a post. That is precisely why emotionally charged content can travel so quickly. The platform rewards immediate action; democratic judgment often requires delay.

Final answer: Are social media algorithms bad for democracy?

Social media algorithms are neither inherently bad nor automatically democratic. They are systems that organize public attention, and their effects depend on what they optimize, how transparent they are, and how much control users and independent institutions have.

The evidence supports several careful conclusions:

• recommendation systems can change what political content people encounter;
• user choices and existing identities remain major influences;
• high engagement can reward emotional or misleading material;
• algorithmic exposure does not always translate into changed political beliefs;
• platforms can widen civic participation and access to information;
• targeted advertising and personalized feeds can make public scrutiny harder;
• false information can spread quickly, especially when it is novel or emotionally charged;
• better evidence requires independent research, platform data, and country-specific study.

The central democratic issue is not whether every person sees the same feed. It is whether citizens can understand how information reaches them, verify important claims, encounter trustworthy alternatives, and participate without being quietly manipulated by systems optimized mainly for attention.

In my view, the healthiest approach combines individual caution with public accountability. Users should slow down before sharing, platforms should provide real control and meaningful explanations, researchers should be able to test claims independently, and election authorities should remain the first source for voting information.

Recommendation systems will continue to shape modern democracy. The public should not have to accept their effects as invisible or unchangeable.

Sources and further reading

• Nature, “How do social media feed algorithms affect attitudes and behavior in the 2020 U.S. presidential election?”
• Nature, “Like-minded sources on Facebook are prevalent but not polarizing”
• Science, “Exposure to ideologically diverse news and opinion on Facebook”
• Science, “The spread of true and false news online”
• Meta Transparency Center, ranking and recommendations
• TikTok, How TikTok recommends videos
• Pew Research Center, Social Media and News Fact Sheet
• Reuters Institute, Digital News Report 2024
• World Health Organization, Infodemic
• Federal Communications Commission, AI-generated voices in robocalls
• European Commission, Digital Services Act
• CISA, Election Security