Dubai's New Deepfake AI SARAAB : Why Digital Trust Matters for Businesses

On September 16, 2026, at Gisec Global at Dubai Exhibition Centre, Expo City Dubai, the Dubai Electronic Security Centre (DESC) unveiled SARAAB, an open-source artificial intelligence model designed to detect deepfake videos. DESC describes it as the first model of its kind developed by a government entity in the Arab region, built entirely by an Emirati team. It is intended to be made available through developer platforms for AI engineers, cybersecurity specialists and researchers to test, use and improve.

SARAAB is a response to a problem that is no longer theoretical. As AI makes synthetic video and audio increasingly convincing, businesses face a newer, less familiar question alongside the usual cybersecurity ones: can the digital content they publish, encounter or rely on actually be trusted, and can their customers tell the difference between the real thing and a convincing fake?

What is SARAAB AI

SARAAB, Arabic for mirage, is built to see past video content that looks real but is not.

• Developed by Dubai Electronic Security Centre. Built entirely in-house by an Emirati engineering team at DESC, rather than commissioned from an external vendor.
• An AI model for detecting deepfake videos. Its specific purpose is identifying manipulated or synthetically generated video content.
• Open source. DESC plans to release SARAAB on Hugging Face, a widely used open-source AI platform, so the underlying code is accessible and modifiable rather than locked behind a commercial license.
• Built for developers, researchers and cybersecurity professionals. The intended audience is technical: AI engineers, security specialists and researchers who can test it, adapt it and contribute improvements, not a consumer-facing app.
• Announced at Gisec Global 2026. Unveiled on September 16, 2026, the opening day of the exhibition at Dubai Exhibition Centre, Expo City Dubai, alongside two companion cybersecurity initiatives.

DESC has said SARAAB will be made available through developer platforms so specialists outside government can test it, stress it against new deepfake techniques, and contribute back to its development. That open, collaborative model matters, because deepfake generation techniques keep changing, and a detection model that cannot evolve alongside them has a short shelf life.

How does SARAAB detect deepfake videos

Based on DESC's own comments to Khaleej Times, SARAAB analyses facial regions within a video using heat-map technology, a method that highlights the specific areas of a frame where manipulation is most likely, rather than making a single pass or a single yes-or-no judgement on the whole clip. Critically, the system is built to examine a video across its full duration rather than checking only one frame or one short segment.

Hassan Majid Alkhazraji, Senior AI Executive at DESC, explained the reasoning in comments reported by Khaleej Times: if a video runs for two minutes, SARAAB analyses the entire two minutes, because the manipulated portion might only be the final 30 seconds. A deepfake does not have to run through an entire clip to do damage, and a detection tool that only samples a frame or two can miss exactly the segment that was altered.

DESC has reported a 91 percent accuracy rate for SARAAB in testing, which it has described as among the highest reported for a deepfake detection model. As with any new detection model, that figure reflects DESC's own testing conditions rather than an independent, third-party benchmark, and public technical documentation on the model is still limited ahead of its full release. It is worth treating the underlying architecture as still emerging rather than fully proven at scale.

Why deepfakes are becoming a business problem

A tool like SARAAB exists because the underlying problem is already real. Deepfake technology has moved from novelty to a genuine operational risk for businesses that publish video content, communicate with customers online, or simply have a public brand identity that someone else could imitate.

Fake brand content

The same AI techniques that make deepfake detection necessary also make it easier to fabricate convincing brand material without a company's knowledge or consent, including promotional videos, executive statements, advertisements, product demonstrations and general social media content. A convincing fake product demo or a fabricated statement attributed to a company executive can spread before anyone at the business is even aware it exists.

Brand impersonation

A fake video that appears to come from a company representative, whether a spokesperson, an executive or a customer service account, can damage customer confidence quickly, particularly if it circulates on social media before the real business has a chance to respond or clarify.

Fraud and social engineering

Deepfake audio and video are increasingly used as part of impersonation attempts, from fabricated voice messages used in payment fraud to fake video calls used to pressure an employee into an unusual request. The believability of the fake is precisely what makes it dangerous, since it can bypass the instinctive skepticism people apply to a suspicious email or an obviously fake message.

Reputation damage

Even when a business has no involvement in a piece of fake content and is not itself the target, customers may still associate it with the brand if the two are visually or contextually linked, at least until the business can demonstrate otherwise. The window between a fake surfacing and a business being able to prove what is real is where most of the reputational damage happens.

What deepfake technology means for digital trust

Digital trust is a broader concept than cybersecurity alone, and deepfakes make that distinction more visible.

Speaking at Khaleej Times' FutureSec forum in Dubai in September 2026, Abdulla Alsuwaidi, director of Cyber Security Planning and Performance at DESC, described digital trust as multidimensional, extending well beyond technical security controls. On a panel titled "Governing the Foundations of a Trusted Digital Economy," Alsuwaidi and fellow panelists framed digital trust as resting on trusted identities, secure transactions, accountability, transparency, human behaviour, and cooperation between government and the private sector, alongside collaboration between national and international entities and a shared sense of responsibility across organisations and individuals. Alsuwaidi also noted that regulation is only one layer of digital trust, and that AI is changing the risks businesses face faster than policy and standards can always keep pace with.

Applied to an individual business rather than a national framework, digital trust comes down to a set of practical questions a customer, partner or search engine effectively asks every time they interact with a company online.

• Is the website genuine. Does it clearly and verifiably belong to the business it claims to represent.
• Is the company information authentic. Are the details published about the business accurate and consistent everywhere they appear.
• Can customers verify the business. Is there a clear, official way to confirm the company is who it says it is.
• Are communications coming from official channels. Can a customer tell an official message from an impersonation attempt.
• Can users safely submit information. Are forms, checkouts and contact channels secured properly.
• Is the brand consistent across digital platforms. Do the website, social profiles and listings all present the same identity.
• Can customers distinguish official content from impersonation. If a fake surfaces, is there enough of an authoritative digital footprint to make the fake obvious.

None of these questions are answered by a deepfake detection model alone. They are answered by how a business builds and maintains its own digital presence, which is where website development, branding and search visibility come in.

How UAE businesses can strengthen digital trust

Building digital trust is less about reacting to any single deepfake incident and more about maintaining a digital presence solid enough that a fake is obviously out of place next to it.

Build a credible business website

A business's website is usually the first place a customer goes to verify it is real. That means a clean, professional interface, clear and accurate company information, HTTPS security, properly secured forms, consistent branding throughout, visible contact information, and clear information about the services or products on offer. A dated, slow or inconsistent website undermines trust just as effectively as an actual impersonation attempt, since visitors read a poor experience as a signal that something is off. See Tomsher's web design and development work for how this gets built in practice.

Maintain consistent digital branding

Trust breaks down quickly when a company's logo, messaging or company description looks different depending on where a customer encounters it. Keeping logos, brand identity, messaging, social profiles, website information and company descriptions consistent across every channel makes it far easier for a customer, or an AI system summarising the business, to recognise what is genuinely theirs. Tomsher's corporate branding work covers exactly this kind of consistency.

Make business information easy to verify

A Google Business Profile, an official website, accurate company contact information, active social profiles, structured data, and consistent business information across all of these are what let a customer, a search engine or an AI assistant confirm a business is genuine in a few seconds rather than a few minutes. This is standard SEO and local SEO territory, but it doubles as verification infrastructure. See Tomsher's SEO services for how that gets structured.

Think beyond traditional SEO

As AI systems increasingly answer questions about companies directly, rather than simply listing links to their websites, businesses need reliable and consistent information across their entire digital footprint, not just their own site. That means structured content, schema markup, clear entity information, authoritative pages, an accurate Google Business Profile, and consistent company information wherever the business is mentioned. This is the practical overlap between SEO, AI search optimisation and generative engine optimisation, and it is increasingly what determines whether an AI assistant describes a business accurately or not at all. See Tomsher's generative engine optimisation work for more on this.

Can AI help businesses fight AI-generated threats

The pattern behind SARAAB is a familiar one in security: AI-generated threats are increasingly met with AI-powered detection, backed by human verification rather than automation alone.

SARAAB is a government-built research and detection tool aimed at specialists, not a product a business installs to protect its own brand, and it is worth being precise about that distinction. For an individual business, the realistic path forward is a combination of layers rather than any single tool.

• ✓AI-based detection tools for flagging suspected synthetic content, as they mature and become accessible beyond research and government use.
• ✓Standard cybersecurity controls covering accounts, communications and infrastructure, since most deepfake-enabled fraud still relies on a conventional security gap somewhere.
• ✓Human verification for anything unusual, particularly requests involving payments, credentials or urgent action, regardless of how convincing the video or voice sounds.
• ✓Trusted, official digital channels that customers and partners already recognise as the business's real presence.
• ✓Brand monitoring to catch impersonation attempts, fake accounts and fake websites early, before they spread.
• ✓Secure, well-built websites that are difficult to spoof convincingly and easy for a customer to confirm as genuine.

Detection technology like SARAAB addresses one layer of this, identifying manipulated video after the fact. The layers above it, a verifiable brand and a secure digital presence, are what a business controls directly, and they matter regardless of how good deepfake detection gets.

What SARAAB means for Dubai's technology sector

SARAAB was not unveiled alone. DESC used the same Gisec Global 2026 platform to launch two companion initiatives: the Dubai Cyber Skills Framework, a standardised reference for cybersecurity qualifications, training and career progression aimed at professionals, employers, academic institutions and training organisations, and the ISR Auditor Certification Programme, which trains auditors and information security professionals to assess compliance with Dubai's Information Security Regulation using consistent evaluation standards.

Taken together, the three initiatives point to a broader pattern rather than a single product launch. Dubai is not only adopting AI technologies at pace, it is also building the tools, standards and certified expertise needed to manage the risks those same technologies introduce. A government-built, open-source deepfake detection model sits naturally alongside a skills framework and an auditor certification programme, since all three are attempts to make a fast-moving technology sector more accountable rather than simply faster.

For businesses operating in Dubai and the wider UAE, that is a useful signal. Deepfake detection, cybersecurity skills standards and information security auditing are moving from ad hoc practice toward recognised infrastructure, which means digital trust is likely to become a more explicit expectation, not just a best practice, over the next few years.

What businesses should do next

None of this requires a business to build its own deepfake detection system. It does mean treating a verifiable digital presence as infrastructure, not an afterthought.

1. Establish official digital channels. A proper website, verified social and listing profiles, and clear official communication channels customers can recognise as genuinely the business's own.
2. Keep brand information consistent. The same company details, description and messaging across the website, social profiles, directories and listings.
3. Strengthen website security. Secure forms, an up-to-date CMS, and properly managed plugins, integrations and access controls.
4. Monitor for brand impersonation. Watch for fake websites, fake social accounts and fabricated content using the company's name or identity.
5. Build an authoritative online presence. SEO, local search, structured data and genuinely useful content that gives a business real digital weight.
6. Educate employees. Particularly sales, marketing and management teams, on how deepfake-enabled fraud and impersonation actually work.

The future of digital trust in the AI era

AI is making digital content easier to create, at a pace that shows no sign of slowing. Technologies such as SARAAB are part of the emerging effort to make that content easier to verify again, closing at least part of the gap AI itself opened. Neither side of that race is finished, and it is unlikely to be for some time.

For businesses, the practical implication is that a trustworthy digital presence will increasingly depend not only on what a company publishes, but on whether customers, partners and AI systems can confidently identify the company's authentic digital identity in the first place. That is precisely where website development, branding, SEO and digital marketing intersect with digital transformation more broadly. A detection model catches a fake after it appears. A well-built, consistent and verifiable digital presence is what makes the fake obvious in the first place.

Related topics this connects to

Frequently asked questions