Introduction

For the humanitarian sector, the 2004 Indian Ocean tsunami marked the first major disaster to receive widespread digital coverage, while the wars in Afghanistan and Iraq (in the early 2000s) became the first armed conflicts dissected in real time by thousands of online commentators. This signalled the start of a decentralized digital information era, shaped less by traditional media and more by fast, participatory online spaces. Early digital content often failed to reflect local realities, particularly in contexts where local languages were absent from platforms. Yet the emerging blogger community played an active role in reporting, verifying content and calling out manipulated imagery,¹ fabricated reporting² or exaggerated harm. While imperfect,³ these efforts exposed new possibilities for scrutiny and public accountability.

These online blogger communities emerged as a powerful force within the information ecosystem. Enabled by technological innovation and driven by social shifts, individuals were able to share unfiltered perspectives, often in response to growing dissatisfaction with mainstream media. Blogs offered immediacy and more personal, alternative narratives, fostering informal networks of discussion (‘communities’) and shaping early forms of participatory media. Ultimately, they laid the groundwork for today’s decentralized and user-driven communication models.

For humanitarian operational and communication professionals, this period felt fast-paced and unpredictable, heralding changes that were once unimaginable. Humanitarian crises – especially wars – have always been accompanied by the use of information to deceive, manipulate, discredit, disrupt or advance political or ideological agendas.⁴ What has changed is the nature, speed, scale and accessibility of harmful information. Today, a wide range of actors – professional and amateur, anonymous and overt – create and spread harmful information across digital and offline channels. The internet, mobile phones, social media platforms and messaging apps serve as powerful amplifiers. In many spaces, fact, truth and accuracy have been displaced by opinion, emotion and perception – often driven by financial, political or ideological motivations. Polarization and distrust are often not just by-products but deliberate objectives. This dynamic extends well beyond warfare: harmful information now shapes how communities understand, prepare for and respond to disasters, health emergencies, migration and other humanitarian challenges.

Instead of facilitating access to trustworthy, life-saving information, the information space increasingly feeds on people’s fear and uncertainty. Crisis information is routinely manipulated, politicized or distorted – including minimizing a disaster’s cause or severity. For example, in 2024, false information spread online claimed that Hurricane Helene in the US was geo-engineered.⁵ Such theories circulated widely on social media, illustrating a growing trend of disaster misinformation and disinformation, and mistrust in official information.

If the Indian Ocean tsunami was the first disaster to be digitally witnessed at scale, COVID-19 was the first to be digitally lived. Every aspect of the crisis – health guidance, rumours, fear and solidarity – spread online as quickly as the virus itself. COVID-19 – an unparalleled health emergency in living memory – combined a pandemic and infodemic,⁶ affecting people worldwide. While the health crisis has largely been contained, the infodemic has left lasting consequences: deepened distrust in authorities, increased vaccine hesitancy and refusal, and the rise of an industry of influencers and conspiracy theorists spreading narratives about the dangers of everything from medical treatments and health care professionals to cell phone towers.⁷

The impact of such information is immediate and often harmful. Correcting it is extraordinarily difficult. Exploiting social divisions, political sensitivities and deep-rooted prejudices, harmful information increasingly drives people to turn against authorities, reject aid and dehumanize or mobilize against those perceived as ‘other’.

What happens if misinformation lands on me now, somehow it breeds fear on me, and now I find it difficult to identify which one makes sense to me, which one I should follow. Because I get all of them from the media still. I have gotten this one which tells me to do this. I›ve also gotten this one which is telling me if you do this, you›re going to suffer from this, you›re going to get this. So I remain there. So it becomes really very, very difficult and confusing”.

During the 2024 Valencia flood response in Spain, misinformation claimed that aid was being diverted to migrants rather than reaching disaster-affected communities, fuelling xenophobic narratives. Similarly, in Southport (UK) in 2024, following the killing and injuring of several children at a dance workshop, false rumours and inflammatory content spread that suggested the attacker was a Muslim asylum seeker. This narrative inflamed anti-immigrant and Islamophobic sentiment, triggering violent protests and riots outside a mosque and sparking wider unrest across multiple cities.

Contributor Insight 1.1

Physical

Psychological

Social

Societal

Informational

Harmful Information and the DANA floods in Valencia (Spain) 2024 (Part 1)

In late October 2024, Valencia was struck by one of Spain’s deadliest floods in recent history, caused by a DANA (Depresión Aislada en Niveles Altos, or ‘cold drop’). Torrential rain fell in just a few hours and overwhelmed infrastructure, leaving 236 people dead and devastating communities.

Specific incidents of harmful information that spread on social media included negative comments, insults and threats directed at our workers and volunteers on the streets and, to a lesser extent, acts of vandalism against our offices and vehicles (such as graffiti and flat tyres).

Misinformation about the Spanish Red Cross has circulated for years, but during the DANA it mainly took the form of hoaxes and rumours claiming that the National Society was:

• not present on the ground helping people in Valencia

• only in Valencia taking staged photos and videos, rather than providing real assistance – «your clothes are clean, your work tools are new and recently purchased» (see Figure 1.1)

• helping migrants instead of Spanish citizens affected by the DANA

• misusing funds – lying about their destination and amounts allocated or refusing in-kind donations to keep the money.

Fig 1.1

Posts on X, November and December 2024

A smaller number of recurring messages attacked the Spanish Red Cross’ international work (e.g., for Ukraine or Gaza) or resurfaced older messages related to migrants in Spain or blood donations.

In this case, harmful information did not directly hinder the humanitarian response; it was not an obstacle to providing assistance. However, it undoubtedly had an impact on the people involved in the response by:

• creating an extraordinary workload dedicated to denying, explaining or minimizing the impact of harmful information

• causing emotional distress and even doubts about the organization, both within society and among Spanish Red Cross members

• undermining public trust, with some donors questioning the organization.

During the DANA, the groups most vulnerable to misinformation were our staff and volunteers (both on the ground and off), our partners and allies, and, of course, our beneficiaries. Migrants were also targeted in the harmful information – these posts inevitably fuelled xenophobia – but in this case, they were used more as a tool to target the Spanish Red Cross. These groups were emotionally affected to varying degrees and we estimate that some also experienced a certain loss of trust in the organization.

María Trénor Alvargonzález

Director of Communications

Spanish Red Cross

Kenan Terzic

Social Media Manager

Spanish Red Cross

These impacts of harmful information are often felt most acutely at the community level, though in some contexts they are being actively fuelled by external actors. This unfolds against a backdrop of declining trust in traditional institutions and growing perceptions of non-governmental organizations (NGOs) as partisan or politically aligned. This erosion of trust is further exacerbated by what has been termed the “death of expertise”,⁸ as journalists, academics and public institutions – once seen as trusted sources of information – are no longer widely trusted in many countries. Meanwhile, AI is being used to produce content that imitates authoritative, expert voices but may be inaccurate, misleading or stripped of its context.

1.1 Defining harmful information

Today’s information ecosystem is highly complex, encompassing many forms of harmful content, including misinformation, disinformation, malinformation, hate speech, propaganda, foreign information manipulation and interference. These forms of harmful information often overlap and reinforce one another, with intent the key factor that distinguishes them.

This report uses the term ‘harmful information’ to focus on its impact and the responses – rather than rigid classifications, which are often politically charged and context dependent. While there is no universally agreed definition, for the purposes of this report, harmful information refers to information that has the potential to cause, contribute to or result in harm to an individual or entity.

Legitimate criticism, even when uncomfortable or challenging, is not considered harmful. However, organizations must still acknowledge and address it, as unaddressed criticism can be exploited by others.

Understanding harm is essential not only to assess its impact on individuals, communities, organizations and societies, but also to design appropriate, proportionate and principled strategies for response.

1.2 The evolving information ecosystem

Harmful information is often framed as an online phenomenon, but humanitarian crises have long been shaped by rumours, myths and propaganda spread through offline means – word of mouth, pamphlets, radio and television broadcasts, community meetings and other official or semi-official channels. These offline dynamics remain deeply influential, especially in contexts with limited digital access or low media literacy. Importantly, harmful information often moves fluidly between online and offline spaces, amplifying its reach and creating real-world consequences for individuals and communities.

The digital information environment – spanning both digital and cyber domains – has become increasingly complex, contested and politicized. States approach this environment from distinct angles. Some focus on infrastructure – the cables: the physical architecture that enables data flows, such as data centres, cloud servers and undersea cables. Others concentrate on content : the narratives, data and discourse circulating within and across platforms. These two lenses – cables and content – are not mutually exclusive, but they shape how influence and control are asserted over the information space, with implications for sovereignty, surveillance and resilience.

Control over content can involve shaping, restricting or distorting information flows for political, ideological or security purposes. This may be achieved through disinformation campaigns, censorship, platform regulation, information manipulation or influence operations. As a result, the boundaries between cybersecurity, information influence operations and geopolitical competition are increasingly blurred. Strategic decisions regarding cables and content have direct consequences for access to information, freedom of expression and communities’ ability to engage safely and meaningfully in the digital space.

At the same time, threat actors’ tactics, techniques and procedures are no longer confined to technical systems. They increasingly leverage harmful information motivated by financial, political or ideological agendas, making information manipulation as consequential as cyberattacks on physical systems. Psychological factors are at the core of this threat: harmful information exploits fear, emotion, identity and grievance, deepening social division and making it far harder to counter with facts alone.

Contributor Insight 1.3

Information first responders: building narrative resilience into humanitarian response

In an age when crises are amplified by misleading information and manipulation, the humanitarian sector must treat the information environment as operational terrain. False narratives no longer circulate on the margins: they now shape perceptions, drive behaviours and determine the legitimacy of aid itself. From manipulated evacuation route guidance to targeted information attacks on humanitarian actors, harmful content is no longer a side effect of crises – it is part of the crisis.

During the early weeks of the COVID-19 pandemic, false cures, conspiracy theories and xenophobic narratives competed strongly with verified public health guidance, often outpacing and overshadowing it in the battle for attention. In more recent crises, from Sudan to Ukraine, misinformation has obstructed safe passage of civilians, fuelled local suspicion toward aid workers and led to violent reprisals. These examples underscore that the costs of harmful information are not abstract – they are immediate and sometimes life-threatening. Waiting to react means being late to the race.

We are experiencing a crisis in our information ecosystem where distorted narratives erode trust, deepen polarization and fracture the relationship between responders and communities. Edelman’s Trust Barometer 2025 found that nearly 70% of global respondents worried about being purposefully misled by authoritative sources such as government or media. An alarming 25% of respondents said they were willing to spread disinformation to achieve change, a number that rises to 34% among young adults.

Meanwhile, recent data from the Reuters Institute shows 58% of global respondents are concerned about how they can distinguish truth from falsehood online. In many regions, online influencers and national politicians are now seen as the leading sources of inaccurate information – surpassing traditional media, foreign powers or rogue actors.

For the humanitarian sector, the message is clear: the fight for trust cannot begin after a crisis erupts. Just as supply chains are pre-positioned, communicators, digital monitoring protocols and trusted local messengers must be identified and deployed in advance, ready to scale up in the event of a crisis, but more importantly an ever-present player on the pitch.

Such ‘information first responders’ are essential to compete in contested digital spaces and inoculate audiences against rapidly spreading falsehoods. This requires mobilizing authentic individuals who are relatable voices within key communities, building multilingual content pipelines and training field teams to identify and counter harmful narratives. It also requires deploying social media monitoring, real-time alert systems and rapid myth-debunking toolkits. Although AI can cause enormous harm by supercharging the generation and dissemination of fake content, it also presents opportunities to provide the superpower to create accurate and engaging content.

Building resilience to manipulation in a polluted information ecosystem means more than getting the facts right. It means designing communication systems that are agile, community led and built for trust at their core. If humanitarian aid is to remain a neutral and stabilizing force, trust must be treated as a mission-critical capability – one that needs resourcing and defending.

Oliver Hayes OBE

Head of Counter-Disinformation

EMEA

Dave Fleet

Head of Global Digital Crisis and Counter-Disinformation

Edelman

Other factors shaping the information ecosystem are discussed in Chapters 2 and 3.

1.3 Who are the threat actors?

A diverse range of threat actors create and amplify harmful information from lone individuals to paid contractors, coordinated troll networks to outsourced call centres, propagandists to inauthentic accounts, national to transnational entities. Each requires a differentiated response. Whether motivated by profit, ideology, coercion or ego, these actors operate within a vast and expanding influence economy. Some act defensively, others offensively, amplifying noise or falsehoods with strategic intent.

The introduction of AI has further lowered the barrier to entry, enabling more sophisticated, scalable manipulation and deepening the asymmetry between those spreading harmful information and those trying to contain it. What most clearly differentiates these actors is their intent – whether to deceive, disrupt, distract, divide or dominate. That intent not only shapes the forms their actions take but also determines the countermeasures required. This is not a level playing field: those spreading harmful content are faster, louder and have fewer constraints than those working to uphold truth and trust.

Fig 1.2

Threat actors involved in harmful information

1.4 Information integrity in crisis situations

The UN has identified strengthening information integrity as “one of the most urgent challenges of our time” – fundamental to human rights, peace and sustainable development. Information integrity refers to the accuracy, consistency and reliability of information. In response to growing threats from misinformation, disinformation, hate speech and the misuse of digital technologies, including AI, the UN Global Principles for Information Integrity⁹ were launched. These principles are:

• Societal Trust and Resilience

• Healthy Incentives

• Public Empowerment

• Independent, Free and Pluralistic Media

• Transparency and Research.

Preceding these principles, a 2023 UN Policy Brief addressed information integrity on digital platforms, calling for follow-up steps including consultations and the possible development of a code of conduct to help “guide Member States, the digital platforms and other stakeholders in their efforts to make the digital space more inclusive and safe for all, while vigorously defending the right to freedom of opinion and expression, and the right to access information”.¹⁰ The consultations ultimately led to the development and launch of the principles. In September 2025, the UN released the first in a new Issue Brief series titled From Principles to Practice: Strengthening Information Integrity.¹¹

1.5 From broadcast to two-way engagement

The promise of new tools for community engagement was meant to mark a broader shift toward more inclusive communication. Digital platforms enabled a move from one-way, broadcast messaging – press releases, websites, official statements – towards two-way engagement. But in practice, this much-valued shift has become increasingly complex. Anyone can now respond to, challenge, reinterpret or amplify a message instantly and publicly.

While two-way engagement invites dialogue, it also opens the door to criticism, harmful content and narrative disruption, including from global audiences and malicious threat actors far removed from the realities on the ground. Social media rarely fosters meaningful exchange, especially when responses spiral far beyond the original message. For humanitarian organizations, committed to neutrality and impartiality, an unpredictable, emotionally charged environment is difficult to monitor and manage, and failure to navigate it can have direct operational consequences.

As of June 2025, platforms such as Google (26 years old), Facebook (21), YouTube (20), Twitter/X (19), VK (18), WhatsApp (16), Instagram (14), Signal (12), Telegram (11) and TikTok (9 in China, 8 globally) reach billions of people worldwide. These platforms have transformed how people access and share information, connect and engage. But they are also powerful vectors for harmful content. Their algorithms prioritize engagement – often the most shocking or polarizing material – because it generates more clicks, attention and ultimately profit. The result is not a glitch but a feature: systems that reinforce bias, create echo chambers, distort perception and deepen division.

The deliberate spread of harmful information erodes trust, casting doubt on humanitarian intentions, principled action and legitimacy. In some cases, it has led directly to threats and violence: refugees and migrants being menaced, humanitarian staff being attacked and volunteers facing hostility.

During the Ebola outbreak in Guinea in July 2014, the President of the Red Cross Society of Guinea warned that violence against volunteers was hindering access to affected communities and disrupting safe burials. These attacks stemmed from misconceptions and fear surrounding Ebola, despite efforts to counter rumours through radio, television and community engagement. By working with religious leaders, families and local stakeholders, the National Society developed respectful burial practices adapted to cultural norms – allowing people to honour the dead.¹²

Social media users increasingly expect rapid, authentic responses. This creates pressure on organizations to engage in real time, often in tension with centralized approval processes, coordination needs with teams on the ground and the imperative for accuracy. For example, during safeguarding or integrity incidents, the demand for swift and transparent communication can clash with the requirement to verify facts, uphold duty of care and ensure due process. Delays or overly cautious messaging may then be perceived as prioritizing organizational reputation or funding relationships.

In an environment where harmful information moves faster than facts and institutional credibility is under constant scrutiny, even well-intentioned engagement can appear reactive, defensive, disingenuous or lacking empathy. Building genuine two-way engagement requires more than responsiveness – it demands intention, transparency and sustained presence. Yet, when malicious actors exploit an organization’s communications, such as using comment sections to disrupt, derail messages or discredit an organization, two-way engagement may not shift the narrative. Many organizations avoid responding for fear of escalation, but silence can leave harmful narratives unchallenged and weaken credibility. Comment sections, once viewed as spaces for engagement, have become increasingly challenging to moderate. The sheer volume of harmful content they attract, combined with limited capacity to manage it, has led some news and media organizations¹³ to disable them to reduce reputational risks and curb the spread of harmful information.

A good example of responding comes from Haiti during the 2010 earthquake. At that time, the IFRC established a Beneficiary Communication Programme that created ten access points for individuals and communities to share feedback on the humanitarian aid they were receiving and express their needs. These access points included live radio programmes, text messaging, information centres, direct engagement with communities and call centres. This initiative enabled the IFRC to continuously adapt its programming and make more informed decisions based on direct community input.

Knowing when – and how – to communicate is increasingly difficult. Delays or silence often create a vacuum that others, whether ill-intentioned or misinformed, will quickly fill.

1.6 Who is most vulnerable to harmful information – and why?

Despite the unprecedented volume of information available today, many people still live in information vacuums where vital, life-saving information is inaccessible or unreliable. Harmful information can affect individuals, communities, institutions and entire societies, but certain groups are disproportionately at risk due to structural, contextual or situational vulnerabilities. As political and security analysts Singer and Brooking observe: “Like any viral infection, information offensives work by targeting the most vulnerable members of a population – in this case, the least informed.”¹⁴

In humanitarian contexts, those in the most vulnerable situations may include:

• Crisis-affected populations: People living through armed conflict, disaster, displacement, migration or health emergencies often face disrupted communication channels, heightened anxiety and scarce or manipulated information.

• Groups that are marginalized or socially excluded: Communities facing discrimination based on race, ethnicity, gender, religion, disability or legal status are frequently targeted by harmful narratives and often lack access to reliable sources of information.

• Youth and older populations: Young people may be more exposed to harmful information due to time spent online, peer influence or online trends. Older populations may struggle with digital literacy and navigating new information environments.

• Humanitarian staff, volunteers and organizations: These are increasingly targeted by narratives questioning their neutrality, independence, impartiality or integrity, undermining trust, access and security.

• Journalists , human rights defenders and civil society leaders: People in these roles are often deliberately discredited, silenced or endangered in polarized environments.

• People with limited access to information or media literacy : Without trusted information channels or critical thinking skills, these individuals are more likely to believe and spread false content.

I don’t think it’s necessarily low-income people or the elderly who are more susceptible. Rather, it’s those who are more easily influenced by their own emotions and by information – this has nothing to do with age. You’ll find that some people are especially opinionated and unwilling to change their views. In some cases, certain ‘leaders’ see their own judgement as superior to others and insist they’re right. Their will, or their subjective descriptions, can drive the spread of information – and they believe it themselves. For example, I might be a 20-year-old rescuer, but if the person is 50 years old, even if I tell them what’s correct and what’s not, they may question me and say, ‘I’ve eaten more salt than you’ve eaten rice; I know better’. So it’s not age that determines susceptibility to misinformation – it’s subjective bias.”

Identifying and understanding these vulnerabilities, along with the specific risks of harmful information in different contexts, is now essential to mitigating their impact as part of humanitarian response. In a survey conducted by IFRC with 400 staff members, 50% reported observing misinformation, disinformation or hate speech affecting humanitarian work , with the most frequently cited issues relating to vaccinations, health and migration.

1.7 What is the impact of harmful information?

Harmful information threatens the lives, safety and dignity of people in humanitarian crises. It can escalate violence during armed conflict and disasters, distort realities on the ground, mislead people about the availability of aid and influence life-saving decisions – such as whether to stay or flee, or accept or reject medical care.

So misinformation is very dangerous – very, very dangerous, not only in the bigger spectrum. It can also be in our families. It can break families. It can confuse and hijack people’s future. It can wreak havoc into people. It can bring about violence. It can cause a lot. It’s a terror of its own, in my opinion.”

UNHCR has documented a wide range of offline harms linked to harmful information in humanitarian contexts, including xenophobia, racism, persecution, violence, killings, forced displacement, trafficking and exploitation, barriers to accessing rights and services, damaged reputation, erosion of trust and legitimacy, diminished ability to protect and support refugees, threat to the physical security of humanitarian workers and decreased donor support.¹⁵

The World Health Organization (WHO) has highlighted the psychological and social harm caused by harmful information, both within communities affected by armed conflict and among humanitarian responders. As noted in a WHO review, “Based on the available evidence, people are feeling mental, social, political and/or economic distress due to misleading and false health-related content on social media during pandemics, health emergencies and humanitarian crises.”¹⁶ This occurs when harmful information discourages people from seeking or accessing humanitarian services and undermines organizations’ ability to deliver and implement effective interventions.

The International Committee of the Red Cross (ICRC) has noted that sharing certain forms of harmful information may violate international law. International humanitarian law imposes “important limits on publishing or sharing certain forms of [harmful information]”,¹⁷ such as that which encourages violations of international humanitarian law, threatening violence which spreads terror among civilians, unduly interfering with humanitarian or medical work, and publishing images of prisoners of war.

Under international human rights law and other rules of international law, advocating hatred that constitutes incitement to hostility, discrimination or violence and inciting genocide, may amount to violations.¹⁸ The UN Special Rapporteur for Freedom of Expression has cautioned that state-sponsored disinformation “has a potent impact on human rights, the rule of law, democratic processes, national sovereignty and geopolitical stability because of the resources and reach of states and because of their ability to simultaneously suppress independent and critical voices in the country so that there can be no challenge to the official narratives.”¹⁹

The IFRC is concerned that some donors have given undue weight to rumours instead of engaging directly and formally with the affected National Red Cross and Red Crescent Society. This has led to opinions being formed based on disinformation rather than verified facts. While these rumours may have served the interests of certain individuals or institutions, they have negatively impacted communities and hindered the ability of National Societies to expand and deliver essential humanitarian services.

Laws, policies and plans underpin all aspects of disaster risk management (DRM), protecting and preparing communities all around the world. Robust legal and policy frameworks are therefore a crucial piece of the puzzle for promoting information integrity and addressing the challenges posed by harmful information in DRM. To respond effectively, humanitarian actors must understand how harmful information disrupts response efforts. A typology of harm is important in building an evidence base that supports efforts to identify, measure and mitigate these effects. Each of the following types of harm can significantly undermine humanitarian action and all must be better understood, monitored and addressed.

1.8 Artificial intelligence and harmful information

People’s opinions are increasingly shaped by systems they neither fully understand nor can meaningfully challenge. Many assume they are seeing the same content as others on their feeds, but this is rarely the case. AI-driven recommendation algorithms curate what content surfaces, often reinforcing emotional reactions and echo chambers. This dynamic subtly but powerfully erodes individual agency in how information is encountered, interpreted and acted on.

The harmful information landscape has rapidly evolved with the proliferation of AI tools. The publication of the first International Scientific Report on the Safety of Advanced AI (commonly referred to as the ‘Bengio Report’) in January 2025 underscored that, if discerning truth from falsehood was already difficult, AI has fundamentally shifted the balance of power in favour of those who control the production and dissemination of information. The report notes that AI reshapes how information is produced, who is empowered or disempowered by it and how it is manipulated. **** AI can generate persuasive, human-like content, rapidly and at scale – despite lacking deep conceptual understanding. As such content is often indistinguishable from human-produced material, people tend to overestimate their ability to detect it, increasing their vulnerability to manipulation. Emotionally charged or personalized content, especially when combined with social media data, can strongly shape perceptions and beliefs.²⁸

The broader impact of AI-driven disinformation is still debated. Some studies suggest its spread and effects may be limited; others warn it may be more concentrated, harder to trace and prone to unintended consequences. What is certain is that the information space is becoming more complex and contested. Although no scientific consensus exists on the full societal impact of false information, its viral potential is well documented. Efforts such as adding watermarks or filtering content show some promise in detecting and mitigating AI-generated manipulation but remain limited. Moreover, interventions must balance curbing manipulation with protecting free expression. Reports point to an increase in the prevalence of AI-generated deepfake content (see box) but overall, scientific evidence remains limited. Anecdotal reports of harm caused by AI-generated fake content is growing, but without better evidence the full scale and impact remains difficult to assess.²⁹

As AI-generated content becomes increasingly indistinguishable from human-created material, detection remains a persistent challenge. Media authentication techniques such as digital watermarks offer some protection but are easily bypassed or removed, especially in high-risk or adversarial environments.³¹

There is no consensus on whether more realistic fake content leads to more effective manipulation, or whether the main barrier is distribution. Some experts argue that the greater challenge lies in spreading fake content at scale – not creating it.³² Research also suggests that ‘cheapfakes’ (simple, less sophisticated manipulations of audiovisual content) may be just as harmful as sophisticated deepfakes, reinforcing the idea that distribution and reach may matter more than quality.³³ While social media platforms deploy moderation, labelling and source credibility checks to limit the spread of manipulated content, these measures raise concerns about free expression. At the same time, research shows that algorithms often “prioritise engagement and virality over accuracy or authenticity of content”, potentially aiding the rapid spread of AI-generated content to manipulate public opinion.³⁴

The UN report Governing AI for Humanity emphasizes that it is “more useful to look at risks from the perspective of vulnerable communities and the commons”.³⁵ It draws on the AI Risk Global Pulse Check, a survey of 348 AI experts that captures emerging AI-related trends and risks. The report stresses, however, that risk management must go beyond simply listing or prioritizing risks. It advocates for framing risks through the lens of vulnerability – shifting the focus from what the risk is (e.g., ‘risk to safety’) to who is at risk, where these risks occur and who should be accountable. This approach highlights the vulnerability of individuals, political systems, society, the economy and the environment. In the context of safety, the report underscores the importance of ensuring reliability and interpretability of AI systems and of assessing and mitigating risks to individual and collective rights, national and international security, and public safety across diverse contexts.³⁶

Access to AI tools also remains uneven, reinforcing existing digital divides. People without reliable connectivity, digital literacy or language inclusion continue to be excluded or misinformed.

What is clear is the need for systematic monitoring.

Fig 1.3

AI Incident Tracker

This tracker illustrates how the severity of AI incidents has changed over time.

1.9 Narratives shape perception, trust and action

Narratives are the frameworks through which people make sense of their experiences and the world around them. They connect the personal to the political, the local to the global. Strong narratives simplify complexity, offering coherence – often at the expense of nuance. Narratives matter. In harmful information campaigns, they often carry historical or cultural weight, with many actors playing a long game.

An event holds power only if people believe it happened. In today’s information environment, perception can outweigh truth: fabricated incidents can generate real consequences, while verified ones may be ignored. Outcomes are increasingly shaped less by facts and more by psychological, political and algorithmic influence. Trust is easily eroded by distortion and distraction. Political discourse often borrows tactics from the use of information in warfare, while armed conflict is shaped by battles over online narratives.³⁷ The risk is that we all become participants when we click, like or share – feeding a wider contest over attention, belief and bias.

In armed conflicts, civilians are often inadvertently drawn into the information dynamics, exposing them to serious risks to their safety – for example, by posting geolocated photos or videos of shelling on social media, relaying updates about territorial control or documenting attacks to seek help or accountability.

These overlapping, often invisible, struggles shape how we perceive the world and act within it. In this ‘network of networks’, there is no neutral ground³⁸ – the objective is influence, control and power. As Singer and Brooking note, influence is achieved not only through force but through shaping how people interpret the world, triggering emotional reactions that drive action and cultivating a sense of shared identity and belonging. Those who succeed are the ones who construct and spread compelling narratives – consistently, repeatedly and globally.³⁹

Effective narratives in the digital landscape rely on simplicity, allowing messages to be absorbed quickly. This is why images resonate so powerfully. Resonance and novelty drive engagement. Yet simplicity is not something the humanitarian sector excels at; some argue the sector has even complicated public understanding of the humanitarian endeavour.⁴⁰

Geopolitical tensions are intensifying polarization within societies and across borders, fuelling division, mistrust and competing narratives that complicate humanitarian action and undermine social cohesion. Returning to the earlier discussion of cables and content, control over both is increasingly treated as a strategic priority, with direct consequences for humanitarian response.

1.10 Controlling accessibility and content moderation

Content moderation, which combines human oversight with automated tools such as fact-checking and content removal, is conducted according to platforms’ own terms of service. In this role, platforms act as gatekeepers, determining what is permissible and facing criticism both for failing to act decisively against harmful content and for over-censoring. Recently, major platforms have scaled back their moderation efforts, driven by economic pressures, reputational concerns and political or ideological influences. Increasingly, they rely on user reporting or content de-ranking as a way to comply with legal requirements.

Although the internet was founded on openness and trust, it remains vulnerable to shutdowns, blackouts and targeted restrictions.

1.11 Analysis and monitoring of harmful information

Analysing harmful information and its risks is essential to understanding the tactics, techniques and procedures used by those who create and disseminate it. Such analysis must consider contextual, historical, political, social, cultural and economic factors. Several analytical frameworks exist, including the 5 Ws⁴² (commonly used by journalists), ABCDE,⁴³ DISARM⁴⁴ and MITRE ATT&CK,⁴⁵ each requiring different levels of training and analytical capacity. This is explored more fully in Chapter 3, on page 103.

Effective analysis also relies on monitoring both the operational and information environments to detect harmful content, while integrating community-based approaches to better understand its dynamics and impact. Many humanitarian organizations are using a range of tools – commercially available or internally developed – that combine automated and manual methods to scrape public data. This is too often with a narrow focus on reputation, rather than assessing how harmful information affects their security, operations and the broader humanitarian context.

The humanitarian sector urgently needs more systematic monitoring and social listening – including sentiment analysis – to detect, understand and frame responses to harmful information. In this regard, monitoring remains fragmented, underfunded and often ad hoc. By incorporating community-based approaches and sentiment analysis into social listening, organizations can detect and track not only what is being said but also how people feel – fearful, angry, mistrustful or supportive – providing deeper insights into shifting perceptions and risks.

Without structured, real-time detection and analysis of both online and offline narratives, humanitarian actors risk being reactive rather than proactive, allowing harmful information to spread unchecked and erode trust before it can be countered. Systematic monitoring, social listening and sentiment analysis, grounded in ethical safeguards and community participation, would allow organizations to detect emerging narratives early, anticipate risks, adapt communications and programmes in real time, and strengthen resilience at both community and institutional levels.

However, there is general acknowledgement that optimal monitoring tools have yet to be deployed, not because of a lack of tools but primarily due to resource constraints. Concerns also persist around the legal and ethical implications of data scraping, which can be perceived as intrusive or akin to surveillance.

Concluding remarks: harmful information is not background noise

Harmful information is not background noise: it actively shapes how people understand crises, who they trust and whether they can access humanitarian assistance and protection. It influences safety and security both directly and indirectly. The struggle over harmful information is as much about cables as it is about content – the infrastructures and narratives that shape access, trust and power.

As the information ecosystem becomes increasingly complex, so too must the capacity to read it, respond to it and protect affected populations, individuals and organizations from its harms. Navigating this ecosystem is now a core part of what it means to act in humanitarian crises. It must inform how responses are designed and implemented, while also driving advocacy for broader systemic change.

The responsibility cannot rest solely with humanitarian organizations. Addressing harmful information requires enhancing resilience, building trust and deepening community engagement. It also demands coordinated, multi-stakeholder action – by governments, technology companies, media, communities and civil society organizations.

Confronting harmful information is not optional. It requires nothing less than systemic change. It is essential to protecting lives, upholding humanitarian principles and ensuring that humanity itself remains the strongest counterforce to manipulation, mistrust and division.

The task now is to ensure that the necessary safeguards, engagement, proximity, trust and resilience are not left to chance or profit, but are deliberately harnessed to protect people and enable principled humanitarian action.

Asks, aims and recommendations

Recommendations

For governments and policy-makers

• Integrate harmful information management into crisis preparedness and response frameworks.

• Invest in early-warning and verification systems to deliver timely, reliable, life-saving information.

• Uphold cyber norms and protect humanitarian organizations and critical infrastructure from malicious ICT use.

• Establish clear legal and policy frameworks that support and protect humanitarian action from interference and harmful information campaigns.

For technology platforms

• Prioritize rapid moderation and fact-checking in humanitarian crises – before, during and after crises – with effective escalation channels for humanitarian organizations.

• Ensure tools function in low-bandwidth, multilingual and resource-constrained contexts.

• Adapt algorithms to reduce amplification of harmful narratives targeting humanitarian organizations, principled humanitarian action and affected populations.

• Report transparently on moderation actions, algorithmic adjustments and impacts in humanitarian contexts.

• Co-design crisis-response protocols with humanitarian actors to ensure interventions are timely, context specific and aligned with internationally recognized rights and ethical standards.

For humanitarian actors

• Treat harmful information management as an operational risk – not just a communications challenge – integrated into humanitarian diplomacy and protection dialogue with states, regional organizations and multilateral forums; and into programmes, risk frameworks and preparedness planning.

• Train staff and volunteers in rumour verification, digital literacy and safe information programmes; share insights across trusted organizations.

• Strengthen monitoring by combining AI-enabled tools with human expertise for real-time detection, mapping of harmful information ecosystems and early warning; foster collaboration across the humanitarian sector to pool resources and capacity.

• Document and analyse harmful information incidents systematically to build an evidence base for policy, adaptation and advocacy.

• Engage communities transparently, co-creating messages with affected communities and reinforcing neutrality, impartiality and independence.

For communities and local leaders

• Develop and sustain local rumour tracking and verification systems.

• Act as trusted intermediaries to strengthen solidarity, resilience and public confidence in humanitarian response.

• Provide feedback loops to humanitarian actors and authorities on trust gaps and unmet concerns.

• Ensure inclusivity by making sure youth, minorities and people with disabilities are represented in community information systems.

• Foster dialogue, peer-to-peer engagement and community-led initiatives to counter harmful narratives and prevent polarization or stigma.

Endnotes