Reflecting on your journey from academia to leading Pindrop, what inspired you to focus on voice security?

While trying to purchase a suit in India, my credit card was declined, and the transaction was marked as fraud. Unable to resolve the issue over the phone, the embarrassment and helplessness of that moment made one thing very clear: there was a painful gap between trust, finances, and voice. 

That incident propelled my PhD research at Georgia Tech, leading to a fascination with the vulnerability of voice channels and the lack of robust security measures, especially as more interactions shifted to phones and voice assistants.

Founding Pindrop was always about solving that trust problem. We saw voice as the next frontier of fraud, and I wanted to bring scientific rigour and real-world solutions to a space that was ripe for innovation.

Pindrop serves major healthcare providers. What unique voice security challenges have you observed in the healthcare sector compared with other industries?

Healthcare is a complex threat landscape. For financial institutions, the effect of fraud is apparent – it results in tangible, monetary losses. But healthcare fraud is trickier: it involves the theft of sensitive patient information, manipulation of medical records, or abuse of access to prescription systems. Plus, patients may call from a range of phone numbers or have family members or caregivers speak on their behalf, making identity verification far more complex.

The healthcare industry has become the most targeted sector for data breaches. According to the [US] National Library of Medicine, over 41.2 million in healthcare records were exposed in 2019 alone, and the US Department of Justice estimates that healthcare fraud may be costing the healthcare system US$100 billion each year. With that in mind, proper caller identification and authentication are paramount to addressing healthcare security challenges and improving the consumer experience. Voice security protocols help identify and authenticate callers, enhance the caller experience, and provide diagnostic insights to reduce potential security threats in the contact centre. 

The healthcare industry has become the most targeted sector for data breaches... proper caller identification and authentication are paramount

Can you share insights from Pindrop’s 2025 Voice Intelligence and Security Report regarding how AI-powered voice attacks are reshaping fraud risk across the healthcare and insurance sectors?

The 2025 report paints a clear and urgent picture: artificial intelligence (AI) powered voice attacks are rapidly increasing and fundamentally changing the fraud landscape, particularly in healthcare and insurance. Insurance fraud alone rose 19% year over year in 2024, with fraud now occurring in one out of every 4,359 calls across life, health, and property casualty lines. This is cause for alarm, in part because average fraud exposure in insurance is 20 times higher than banking. That means a successful attack could be even more financially devastating.

We’re also seeing a shift in attacker behaviour. Unlike other industries where phishing and reconnaissance dominate, only 23% of insurance fraud cases fall into those categories. Instead, fraudsters are leveraging social engineering, procedural knowledge, and ‘familiar fraud’, where attackers exploit personal connections, which now accounts for 7% of fraudulent insurance calls. They’re analysing how systems work and weaponising AI to scale their efforts, often with synthetic voices and multiple touch points.

Voice security must go beyond static authentication. The way forward is a multi-layered defence, with behaviour analysis, liveness detection, and real-time risk signals working together to catch fraudsters quickly.

The report features a real-life case of a fraud ring targeting a major US west coast insurer. What specific cybersecurity threats did this incident reveal, and what steps can insurers take to mitigate similar risks in the future?

It’s clear from that case: today’s attackers are blending persistence, insider-level knowledge, and AI-driven tactics to breach even the most protected systems. 

The fraud ring used multiple calls from offshore numbers to systematically test social security numbers, bypass knowledge-based authentication (KBA) with confidence, and ask strategically crafted questions about policy loans and annuity withdrawals, demonstrating not only access to sensitive data but also an in-depth understanding of internal procedures.

Today’s attackers are blending persistence, insider-level knowledge, and AI-driven tactics to breach even the most protected systems

To mitigate these threats, insurers must stop treating voice as a secondary channel. Instead, they need to adopt layered voice security approaches such as real-time risk analytics, AI-driven liveness detection, and behavioural voice analysis that can detect not just who’s calling, but whether the behaviour and signals match legitimate intent. It’s about moving from static to adaptive, intelligence-driven defences.

With the rise of deepfake technologies, how is Pindrop addressing potential threats to patient data and healthcare fraud?

Deepfake audio presents a growing threat in healthcare, where voice is often used to communicate sensitive information and make critical decisions. Attackers are using synthetic voices to impersonate doctors, administrators, or family members – stealing patient data, issuing fake instructions, or even attempting extortion. This erodes trust and poses serious risks to both operations and privacy. Pindrop combats these threats with advanced voice authentication and deepfake detection. 

Our liveness technology analyses thousands of micro-signals – like unnatural modulations or timing irregularities – to distinguish real human voices from synthetic ones in real time. Combined with behavioural analytics and multifactor verification, we help healthcare organisations identify and block deepfake attacks before damage is done.

What are the benefits and challenges of implementing voice biometrics in healthcare settings, especially regarding patient consent and data privacy?

Voice analysis offers a seamless, low-friction way to verify identity, which is especially useful for patients with disabilities or those with limited access to digital devices. However, healthcare organisations must be extremely thoughtful about consent, transparency, and data stewardship. Patients need to know how their caller profiles are used and stored. We work closely with our customers on compliance and other regulations, and implement opt-in frameworks that build trust, not erode it.

How does Pindrop’s liveness detection differentiate between genuine human voices and synthetic ones in real time?

During a call, Pindrop technology analyses audio in real time to determine if the voice is human or machine-generated. It evaluates more than 1,500 subtle voice characteristics like timing and frequency inconsistencies that are nearly impossible for synthetic speech to replicate naturally. By identifying machine-generated audio, Pindrop can detect deepfakes and spoofing attempts at the point of entry, helping businesses maintain trust and security in their voice channels without adding friction for genuine customers.

What future developments do you foresee in voice security to combat emerging threats in the healthcare industry?

Deepfake attacks aren’t just a possibility; they’re a reality – and they’re only growing. These attacks are a pervasive threat, especially in healthcare, where voice is the bedrock of many critical interactions. At Pindrop, we’re building for this reality: enhancing liveness detection, integrating cross-channel intelligence, and designing voice security that anticipates – and outpaces – the next wave of AI-driven threats.