Seeing is not believing: The race to stop AI deepfakes from inheriting the Earth is on
Seeing isn’t believing anymore. As artificial intelligence (AI) makes access to digital cloning, audio and video manipulation tools, and even technologies to forge eKYC commonplace, the world is bracing for the silent siege of deepfake reaching epidemic proportions.
India’s national cyber security agency CERT-IN has already issued an alert on deepfakes this winter with a ‘high’ severity rating, with the potential to spread false information, financial loss and damage to reputation.
In the past week, social media has been agog with fear of AI-generated images now being able to forge anything from restaurant receipts to even Aadhar cards.
Chennai Police on Wednesday warned citizens of the dangers of the Ghibli craze, whereby people were uploading their photos to be transformed into stylised animations imitating the popular Ghibli aesthetic. The danger here? Unofficial websites and digital platforms which offer free downloads or customisation could infect their device with malware, ransomware or viruses, which might lead to identity theft or data loss.
“Deepfakes have moved from novelty to national threat in record time. In just the past year, we’ve seen scammers impersonate CEOs to siphon off crores, job interviews conducted with AI-generated identities, and political misinformation campaigns that manipulated entire electorates. The threat is global, real-time, and dangerously scalable,” said Ankush Tiwari, founder and CEO of pi-labs, which has come up with tools for AI-driven deepfake detection.
If some estimates are to be believed, deepfake scams have skyrocketed in the financial sector. Last year, total global financial fraud losses were nearly 500 billion dollars. Worse, deepfake-related fraud surged by 1000 per cent (ten times) last year, making it one of the fastest-growing threats in digital finance.
The accessibility of AI-powered forgery tools is worsening the crisis — fraudsters today have access to over 2,000 deepfake video manipulation tools, more than 1,000 voice cloning tools and at least 47 known tools specifically designed to bypass KYC protocols.
Worse, most of today’s firewalls and cyber security tools employed by banks and companies may be inadequate to deal with this AI-driven new surge. Result? eKYC fraud is expected to more than double by 2030, posing a severe risk to banks and financial institutions.
“We need AI to beat the menace of AI,” said Tiwari of pi-labs, which has come up with Authentify, an advanced AI-powered deepfake detection solution for the financial sector. “Financial losses are mounting globally to billions of dollars. But beyond money, the erosion of public trust and reputational credibility is perhaps even more costly.”
Governments of the world have taken note, with the US Defence Department running a Semantic Forensics Program, Singapore has an AI Verify Foundation, while the EU has been crowdsourcing solutions with its Deepfake Detection Challenge. India’s parliament was told last week that the home ministry has established the Cyber Crime Coordination Centre to coordinate with authorities and deal with cybercrimes in a comprehensive manner, while the IT Act in its last modified iteration of 2021 will be the operative platform to protect against “unlawful activities on the internet.”
Sci/Tech