Yet, as AI systems grow in capability, a troubling trend has emerged: the rise of malicious AI prompting.

This includes prompt injection, AI jailbreaking, and adversarial manipulation, techniques that exploit not software flaws, but the linguistic and interpretive nature of AI models. It represents a new frontier of cyber risk.

The scale of this challenge is significant. A 2025 report by the UK’s National Cyber Security Centre noted a 30 percent rise in AI-related cyber incidents within a year, with prompt-based attacks among the fastest-growing threats.

Similarly, the World Economic Forum’s 2024 Global Risks Report ranked AI-driven misinformation, manipulation, and system compromise among the top technological risks. These trends confirm that malicious prompting is no longer theoretical, it is an urgent global concern.

An Expanding Attack Surface

The rapid adoption of generative AI tools has dramatically expanded the cyber threat landscape. Unlike traditional attacks that target code vulnerabilities, malicious prompting manipulates how AI systems interpret instructions.

By crafting deceptive inputs, attackers can cause models to ignore safeguards, reveal sensitive information, or generate harmful outputs.

This makes the threat uniquely dangerous. It bypasses conventional security barriers and engages directly with the AI’s reasoning processes.

Research from Stanford University in 2024 found that over 60 percent of tested AI models could be induced to violate safety protocols through carefully designed prompts.

Meanwhile, studies from MIT showed that multi-turn conversations, where malicious intent is gradually introduced, can evade even advanced guardrails.

This evolution highlights a key reality: malicious prompting is not only technical but psychological. It exploits creativity, deception, and human-like reasoning, making defence far more complex.

Limits of Existing Guardrails

AI developers have introduced safety layers, filters, and refusal mechanisms to mitigate risks. However, these protections are not foolproof. Language is inherently flexible, allowing malicious intent to be disguised through metaphors, fictional scenarios, or indirect instructions.

A 2025 European Union cybersecurity audit found that more than 40 percent of tested AI systems could be tricked into generating restricted content through indirect prompts. This demonstrates that current safeguards, while important, are insufficient on their own.

The Need for Layered Security

Addressing malicious prompting requires a multi-layered approach. At the model level, developers must train systems using adversarial datasets to better recognise manipulation attempts. Reinforcement learning processes should be continuously updated to strengthen refusal behaviours, especially when prompts are ambiguous or deceptive.

However, relying solely on model-level defences is risky. Infrastructure safeguards are equally critical.

Zero-trust architectures, where no input is automatically trusted, can reduce the likelihood of harmful outputs causing real-world damage. Additional measures such as strict access controls, sandboxed environments, and output verification systems provide further protection.

Separating user inputs from system-level instructions is also essential. Many prompt injection attacks succeed because AI systems process external data, emails, documents, or web content, without proper filtering. Strengthening these boundaries can significantly reduce exposure to indirect manipulation.

Monitoring and Behavioural Analytics

Continuous monitoring plays a vital role in AI security. Behavioural analytics tools can detect suspicious patterns such as repeated probing, conflicting instructions, or attempts to bypass safeguards.

These systems act as early warning mechanisms, enabling organisations to respond before threats escalate.

According to a 2024 Gartner report, by 2027 more than 70 percent of large enterprises will adopt AI-driven monitoring tools to detect prompt-based attacks. This reflects a growing understanding that AI security is not static, it requires ongoing vigilance and adaptation.

The Human Factor

Technology alone cannot solve this problem. Human behaviour remains a critical vulnerability. Employees often expose systems to risk by entering sensitive information, poorly structured prompts, or unclear instructions. A 2025 Deloitte survey found that nearly half of AI-related security incidents stem from user error.

Organisations must invest in training to promote responsible AI use. Staff should understand what data can be shared, how to structure prompts, and how to identify manipulation attempts. Without this human-centred approach, even the most advanced safeguards will be ineffective.

The Role of Regulation

Policymakers have a crucial role in addressing this emerging threat. Governments must establish clear standards for AI deployment, including data governance, transparency, and risk management. Mandatory audits and safety assessments should become standard practice for high-impact AI systems.

The United Kingdom and European Union have already made progress. The UK AI Safety Institute focuses on evaluating emerging risks, while the EU AI Act introduces strict requirements for high-risk applications. However, global coordination is essential to ensure consistent and effective regulation.

Ethical Leadership and Collective Responsibility

Beyond technology and policy, there is a moral dimension to AI security. The misuse of AI, through manipulation, misinformation, or malicious prompting, undermines trust and social cohesion. Leaders across sectors must advocate for ethical AI use that protects individuals and promotes the common good.

Addressing malicious prompting requires collective effort. Developers, policymakers, businesses, and users must work together to build resilient and trustworthy AI systems. Transparency, accountability, and collaboration are essential to this process.

Conclusion

Malicious AI prompting is one of the defining cybersecurity challenges of our time. As AI systems become more powerful, the risks associated with their misuse will continue to grow. The future of artificial intelligence will depend not only on innovation but on our ability to safeguard these systems.

By adopting layered security strategies, strengthening governance, investing in human awareness, and promoting ethical leadership, we can mitigate these risks. With decisive action, AI can remain a force for innovation, resilience, and human progress in an increasingly complex world.

The decision follows cases within government circles over what they describe as a “pattern of negligence” by social media firms in responding to police requests. Communications Minister Fahmi Fadzil criticised TikTok’s response times.

“TikTok was very slow in providing information… to the point that I had to call TikTok CEO Shou Zi Chew to inform him, ‘this is a crime that’s being committed and your organisation is very slow’,” Fahmi said, warning that such behaviour would not be tolerated.

The trigger for this confrontation was a viral TikTok video in which a man falsely claimed to be a pathologist working on the investigation into the death of Zara Qairina Mahathir, a case that has attracted public attention. Authorities say TikTok’s delay in handling the matter forced the minister to personally intervene.

Top executives of TikTok are expected to appear at Malaysia’s federal police headquarters, Bukit Aman, on Thursday. The Inspector-General of Police and the Attorney-General will also attend the meeting.

Meta has not been spared as the company, which owns Facebook, Instagram, and WhatsApp, is being summoned over disturbing materials linked to paedophilia that spread across its platforms, including content uncovered during a cybercrime operation known as Operation Pedo. 

Authorities have specifically flagged an online group called Geng Budak Sekolah, which circulated indecent content targeting children.

The Malaysian government has classified several categories of online activity as harmful. These include gambling, scams, child pornography and grooming, cyberbullying, and content linked to race, religion, and royalty. Officials argue that these categories pose both social and national security risks.

Fahmi has insisted that every platform must comply with local laws and respond quickly to enforcement requests. “We see these platforms are not taking the matter seriously, so the dialogue process will continue, and we will stress that Malaysian law applies to them and they must comply. We will summon every platform,” he said.

At the Al Grand Prix Conference 2025, Fahmi also disclosed that Malaysia is considering mandatory identity verification for all online sales and advertising. The proposal is intended to limit fake accounts, deepfakes, and fraud. Singapore already enforces a similar policy, and Malaysia is positioning itself to follow that model.

The issue aligns with international trends where Governments from India to Indonesia, and even within the European Union, are tightening regulations on global tech giants, imposing fines, and in some cases threatening outright bans for non-compliance.

Neither TikTok nor Meta has issued an immediate public response to Malaysia’s latest move.

The spread of false information, fake news, and propaganda online has the potential to manipulate public opinion, undermine democratic processes, and sow social discord.

As the boundaries between truth and falsehood blur in the online space, determining responsibility for preventing and responding to disinformation becomes a critical task.

In this context, it is essential to explore strategies for achieving the best outcomes in combatting disinformation and examine examples from Global South Nations where innovative initiatives are being implemented.

In Cyber 2024, responsibility for preventing and responding to disinformation lies with a combination of government, technology companies, media organizations, and individual citizens.

1. Government:

Governments play a crucial role in setting policies and regulations to combat disinformation. They can implement laws to hold perpetrators of disinformation accountable, work with technology companies to develop tools and strategies to detect and remove false information, and fund public education campaigns to promote media literacy and critical thinking skills.

2. Technology companies:

Platforms like social media sites and search engines have a responsibility to enforce their terms of service and combat the spread of disinformation on their platforms.

They can utilize algorithms and artificial intelligence to detect and remove fake news, work with fact-checking organizations to verify information, and provide transparency about how their platforms are being used to spread false information.

3. Media organizations:

Traditional media outlets also play a crucial role in preventing disinformation by producing accurate and reliable news content.

They can fact-check information before publishing it, provide context and analysis to help readers understand complex issues, and promote journalistic standards of accuracy and impartiality.

4. Individual citizens:

Ultimately, individual citizens have a responsibility to be critical consumers of information and not to share unverified or false information online.

By practicing media literacy skills, fact-checking information before sharing it, and engaging in civil discourse, individuals can help prevent the spread of disinformation in the digital age.

Certainly, preventing and responding to disinformation in Cyber 2024 requires a collaborative effort from all stakeholders – governments, technology companies, media organizations, and individual citizens – working together to combat this growing threat to democracy and public trust.

Essentially, strategy for achieving the best outcomes in preventing and responding to disinformation in Cyber 2024 should be multi-faceted and involve a combination of technological solutions, regulatory measures, media literacy initiatives, and international cooperation. Here are some key components of such a strategy:

1. Collaboration between governments, technology companies, and civil society:

All stakeholders must work together to develop and implement coordinated strategies to combat disinformation.

This could involve sharing information and resources, collaborating on research and development of tools to detect fake news, and establishing clear communication channels for reporting and responding to disinformation incidents.

2. Strengthening regulations and enforcement:

Governments should enact strong laws and regulations to hold individuals and organizations accountable for spreading disinformation.

This could include penalties for those who knowingly spread false information, requirements for platforms to be transparent about their content moderation practices, and measures to ensure the independence and credibility of fact-checking organizations.

3. Investing in media literacy and education:

Promoting media literacy and critical thinking skills among citizens is essential for reducing the impact of disinformation.

This could involve developing educational programs in schools, universities, and communities to teach people how to evaluate the credibility of sources, spot misinformation, and engage in respectful dialogue online.

4. Enhancing technological solutions:

Technology companies should continue to invest in developing tools and algorithms to detect and remove fake news from their platforms.

This could include using artificial intelligence to identify patterns of disinformation, collaborating with fact-checking organizations to verify information, and providing users with tools to report misleading content.

There is a growing body of research and evidence to support the effectiveness of these approaches in combating disinformation. For example, studies have shown that media literacy programs can improve individuals’ ability to discern between accurate and false information online.

Additionally, regulatory measures such as the EU’s Code of Practice on Disinformation have led to increased transparency and accountability among tech companies in addressing disinformation.

By implementing a comprehensive strategy that combines these elements, stakeholders can work together to combat disinformation effectively and protect the integrity of democratic processes in the digital age.

Yes, there are examples from Global South Nations of initiatives and strategies aimed at preventing and responding to disinformation. Some of these examples include:

1. India’s Fact-Checking Platforms:

India has a vibrant ecosystem of fact-checking organizations such as Alt News, Boom Live, and FactChecker.in.

These platforms play a crucial role in debunking misinformation circulating on social media and other online platforms, especially during elections and other critical events. They provide valuable resources to the public to verify and cross-check information before sharing it.

2. Kenya’s Citizens’ Journalism Network:

In Kenya, the Ushahidi platform enables citizens to report and verify incidents of misinformation, hate speech, and fake news.

This crowdsourced approach allows for real-time monitoring of disinformation and enables rapid response measures to counter false narratives.

3. Brazil’s Observatory for the Fight against Fake News:

Brazil has established an observatory dedicated to monitoring and combating fake news and disinformation.

The initiative involves collaboration between government agencies, civil society organizations, and academic institutions to track and analyze misinformation trends and develop strategies to counter them.

4. South Africa’s Media Monitoring Africa:

Media Monitoring Africa is an organization that works to promote media literacy and critical thinking skills among citizens in South Africa.

They conduct research, advocacy, and educational programs aimed at empowering individuals to navigate the digital media landscape and identify and combat disinformation.

5. Nigeria is taking steps to combat disinformation and fake news:

The country has established agencies such as the National Information Technology Development Agency (NITDA) and the National Orientation Agency (NOA) to address the spread of fake news and disinformation.

Additionally, the Nigerian government has introduced the Anti-Social Media Bill to regulate the use of social media platforms and prevent the dissemination of false information.

Nigeria has also launched media literacy and digital literacy campaigns to educate the public on how to identify and combat misinformation.

These examples from Global South Nations demonstrate the importance of local initiatives and collaborations in addressing the challenges of disinformation. By leveraging technological solutions, regulatory measures, media literacy programs, and international partnerships, countries in the Global South can effectively prevent and respond to disinformation in their respective contexts.

In conclusion, as we navigate the complex terrain of disinformation in Cyber 2024, it is evident that addressing this multifaceted challenge requires a collective effort from various stakeholders. Governments, tech companies, civil society organizations, and individuals all have a role to play in preventing and responding to disinformation effectively.

By implementing a combination of technological solutions, regulatory measures, media literacy programs, and international collaborations, societies can work towards creating a more informed and resilient digital environment.

Drawing inspiration from examples in Global South Nations, where innovative approaches are being adopted to confront disinformation, we can learn valuable lessons for shaping a more truthful and trustworthy online world.

By recognizing the shared responsibility we all have in combating disinformation, we can strive towards achieving the best outcomes and upholding the integrity of information in the digital age.

[Featured Image Credit]

===

The Writer, Prof. Ojo Emmanuel Ademola is the first Nigerian Professor of Cyber Security and Information Technology Management, and the first Professor of African descent to be awarded a Chartered Manager Status.