AI offers unprecedented opportunities for businesses - from accelerating innovation to creating hyper-personalised customer experiences. Yet its adoption depends on trust: an increasingly essential currency for businesses. PwC’s 28th Annual Global CEO Survey revealed that only 34% of Asia-Pacific CEOs fully trust AI in their core operations, while just half of global consumers feel confident about AI-driven interactions.
Why does this matter? Without trust, the true potential of AI remains out of reach because adoption hinges on confidence, and reputational risks begin to overshadow benefits.
The risks associated with AI range from algorithmic bias and privacy violations to security threats. Australia's Voluntary AI Safety Standard describes these in detail.
Among these risks, AI-generated misinformation and deepfakes are particularly damaging because they can rapidly erode public trust and business reputation. Sumsub’s 2024 Identity Fraud Report shows deepfake-driven fraud quadrupled globally in 2024, with regional surges of 142% in Europe, 171% in US and Canada, 194% in APAC, 255% across Latin America and the Caribbean, 393% in Africa, and 643% in the Middle East.
Real-world cases already show how costly and damaging deepfakes can be. In 2024, a multinational engineering firm's Hong Kong branch faced a US$25m loss when scammers used deepfakes of the company's CFO and managers during a video conference. This incident marks one of the first multi-person deepfake scams recorded in the region. And earlier, in 2019, a deepfake video purporting to show a major social media platform’s CEO making ominous claims about user data went viral, severely damaging the platform’s brand reputation and heightening regulatory scrutiny.
And this is happening at a time when the very nature of trust is evolving. As thought leaders like Rachel Botsman have observed, we are moving away from centralised, institutional trust - where governments, media, and major organisations are the default arbiters of truth - and toward distributed trust. People increasingly rely on peers, platforms and decentralised systems to decide what’s real. AI is accelerating this shift. With its ability to generate convincing yet synthetic content at scale, it undermines the credibility of authoritative sources and adds friction to already eroding trust models. In a digital world where anyone can create content that appears legitimate, the absence of a central source of truth makes it even harder for consumers to navigate what’s trustworthy.
Global regulators are quickly responding to these emerging risks. The European Union's (EU) landmark AI Act introduces strict requirements and heavy penalties for certain forms of AI use cases, effective from August 2024. Similar initiatives in some US states, Canada and Japan show signs of a shift towards tighter oversight, although unlikely to be as stringent as what we see in the EU.
There are no 'black letter' laws for AI in Australia, and while defamation, copyright, and consumer protection laws may offer partial remedies, they weren’t designed for the speed, volume and anonymity of AI-generated synthetic content. Where it’s possible, legal action is slow and constrained, while the damage from viral misinformation can happen in real time. This lag leaves businesses exposed - facing a threat landscape that evolves faster than regulatory responses can keep up. It’s like a Distributed Denial of Service (DDoS) attack - overwhelming your reputation system with harmful content faster than you can respond.
To mitigate these risks, businesses will need to increasingly rely on more than legal protections alone.
AI detection tools offer scalable, automated analysis of synthetic media - spotting clues like unnatural pixel patterns or voice anomalies. But this is a cat-and-mouse game. Generative models continue to improve, eventually outpacing the very tools built to catch them. A 2025 CSIRO research study found that when leading detectors are hit with deepfakes produced by tools they weren’t trained on, their accuracy can collapse to below 50% - worse than a coin-flip. False positives and false negatives are both risks, and these tools require ongoing training and significant technical investment.
Centralised content labelling, as seen on major social platforms, adds a human layer to the mix. Labels like ‘disputed’ or ‘AI-generated’ help users pause before sharing questionable content. It’s simple and familiar but it’s not scalable. Manual labelling struggles to keep up with the volume of content, and public trust in label impartiality can vary, especially during polarising events.
To change the dynamic, platform providers, publishers and large businesses are turning to solutions that establish authenticity from the outset. For example, LinkedIn implemented the C2PA standard to display content credentials on images and videos, allowing users to verify their origin and whether AI tools were used.
Blockchain-based content provenance offers robust transparency by creating immutable records tracing digital content from origin to distribution. This technology ensures traceability and builds confidence in what’s real. However, it verifies origin - not truth - and integrating it into existing systems is not always straightforward.
Crowdsourced verification initiatives, like Wikipedia or X’s Community Notes, harness collective expertise to evaluate content. These platforms can deliver cost-effective, scalable oversight, but depend heavily on active participation and can be skewed by biases or coordinated manipulation.
To counter problems of manipulations and biases, there are emerging approaches that introduce more ‘skin in the game’ for the content authors. Incentive-driven crowd labelling platforms let creators stake a financial bond on their content to promote responsible authorship with the possibility of financial loss when their rigor for truth reporting has been found wanting.
No single method offers a complete solution. Businesses should therefore adopt a layered approach, blending short-term immediate actions with scalable, innovative long-term investments. Tailoring these strategies to your organisation’s unique context, capabilities, and risk profile ensures robust defense against misinformation and deepfakes.
There’s no magic bullet to tackle AI-generated misinformation. Businesses can combine quick fixes, like detection tools, with big-picture solutions that foster trust in digital content. The key is to adapt these strategies to what works best for your organisation based on your risk appetite and level of exposure.
In collaboration with UNSW Sydney, we explored how organisations can move from reactive misinformation controls to a system built on verifiable authenticity. Together, we designed and tested a novel incentive model for content trust, incorporating elements of blockchain, smart contracts and decentralised identity.
At the heart of the model is the concept of a ‘veracity bond’ - a financial stake that content creators attach to their work to signal confidence in its accuracy1. “This model is inspired by Charlie Munger’s aphorism of aligning the incentive to the desired outcome, which in this case, is truth seeking” emphasises UNSW Business School’s Eric Lim. Building on this foundation, we propose the idea of enabling meaningful debate through readers being able to stake a “counter-veracity bond” to challenge content in a contest for truth. Once the content is challenged, a jury of impartial participants evaluates the dispute. If the challenge is upheld, the bond is forfeited and redistributed, and the same applies if the ruling goes against the challenger. There is balance to both sides of the dispute in the quest for truth. Jurors are incentivised to participate through compensation linked to their decision quality, and their own reputations are maintained through peer reviews.
Counter-veracity bonds create a shared sense of responsibility by also making content challengers accountable. When all parties have skin in the game, participation becomes more honest.
Our joint research found that this approach can significantly improve perceived content credibility, increase user perceptions of fairness and procedural transparency, and can encourage creators to take more care in publishing accurate information. “Unfortunately, most solutions are primarily focused on consumers, who are already overwhelmed by a relentless stream of information from platforms built for passive, endless consumption”, notes UNSW Business School’s Sam Kirshner. “Real change requires solutions at the platform level, where the architecture of information flow is designed with accountability, transparency, and shared responsibility in mind.”
This system was designed to rebalance incentives away from engagement-at-all-costs and toward truthful content creation. For businesses that run content platforms or publish content, it demonstrates the viability of an innovative method for ‘built-in’ trust, and an interesting alternative to traditional strategies.
Misinformation moves fast. Trust takes time. Here’s how to stay ahead:
Organisations that move from passive detection to proactive trust-building will lead in the AI age. Are you ready?
UNSW Business School and PwC Australia share a commitment to meaningful collaboration between academia and industry.
Our joint research examines how artificial intelligence, decentralised trust systems, and new models of community incentives are reshaping trust, governance, and organisational decision-making.
We thank Prof Eric Lim and Prof Sam Kirshner for their leadership and contributions to this work in collaboration with us.