StarTalk Radio

Deepfakes and the War on Truth with Bogdan Botezatu

October 17, 2025

Key Takeaways Copied to clipboard!

  • Global scams are estimated to inflict approximately $1 trillion in losses for 2024, a figure likely conservative due to underreporting. 
  • Scammers prioritize immersive communication methods like instant messaging and phone calls to apply pressure, while advanced technologies like deepfakes are used for one-to-many scams targeting trusted public figures. 
  • The psychological damage from scams, particularly romance scams like 'pig butchering,' can be more impactful to victims than the financial loss, and reporting crimes is crucial for law enforcement to accurately assess the phenomenon. 

Segments

Introduction and Scam Scale
Copied to clipboard!
(00:01:00)
  • Key Takeaway: Scams are a trillion-dollar industry globally, exacerbated by AI and cybercrime advancements.
  • Summary: The scale of scams is massive, with the Global Anti-Scam Alliance (GASA) estimating $1 trillion in losses for 2024. This figure is considered conservative because most scams go unreported due to victim shame or lack of central aggregation. The global cybercrime market is estimated at $9 trillion, suggesting scamming is a significant fraction of this total.
Preferred Scam Attack Avenues
Copied to clipboard!
(00:08:12)
  • Key Takeaway: Scammers prefer immersive communication like instant messaging and phone calls over email to apply pressure and urgency.
  • Summary: Hackers favor instant messaging and direct phone calls because they are immersive and allow for immediate pressure tactics. Email is considered a more static means of communication where victims have time to pause and reconsider. Urgent messages, such as those claiming an account is being depleted, prompt faster victim compliance.
Deepfakes in Mass Scams
Copied to clipboard!
(00:13:32)
  • Key Takeaway: Deepfakes are prevalent in one-to-many scams, using trusted figures to promote fraudulent investments or products via compromised large channels.
  • Summary: Cybercriminals train algorithms on footage of trusted influencers, politicians, or doctors to create deepfakes promoting scams like crypto investments. These deepfakes are broadcast on stolen, high-subscriber YouTube accounts or boosted social media ads to reach massive audiences. The scam relies on the trust associated with the impersonated figure to drive victims to a call to action.
Voice Biometrics and Authentication
Copied to clipboard!
(00:10:41)
  • Key Takeaway: A simple recorded ‘yes’ from a phone call can potentially be used by threat actors to bypass authentication or confirm contractual agreements, as voice acts as biometrics.
  • Summary: Silence on a call can be due to technical glitches in spoofing software, or it might be part of a data-gathering effort. Threat actors may be building databases of voice confirmations, where saying ‘yes’ can substitute for a signature. A short conversation can provide enough voice data to spoof an individual’s voice for extended periods.
Scam Psychology and Structure
Copied to clipboard!
(00:30:16)
  • Key Takeaway: Approximately 90% of scams rely on psychology—hacking the brain by triggering emotions—while technology enables wider reach and effectiveness.
  • Summary: Scams exploit human nature through cues like curiosity (failed package delivery) or greed (get-rich-quick schemes). Technology, including real-time translation via APIs, allows scammers to effectively target diverse geographic markets previously inaccessible to them. Victims who are lonely may suffer greater psychological damage than financial loss, sometimes prioritizing the connection over recovered funds.
Defining Scam Terminology
Copied to clipboard!
(00:35:10)
  • Key Takeaway: ‘Pig butchering’ involves gaining a victim’s long-term trust before inflicting massive financial fraud, while a ‘honeypot’ is a defensive tool used by researchers to study criminal tactics.
  • Summary: Pig butchering is named for fattening the pig (victim) before sacrifice; scammers build trust over weeks or months, often through romance angles, before pushing fraudulent investments. A honeypot is a decoy system used by cybersecurity researchers to attract and record cybercriminals’ actions, helping to decompose attack methods and develop proactive defenses.
Defending Against AI Content
Copied to clipboard!
(00:46:43)
  • Key Takeaway: Defense against deepfakes should focus less on technical artifacts and more on contextual likelihood, such as whether a public figure would promote a specific product or use certain language.
  • Summary: As AI technology evolves, technical giveaways like poor lip-syncing become less reliable indicators of a deepfake. A more effective defense involves assessing the content against the known behavior, expertise, and communication style of the person being impersonated. Security solutions are evolving to help users identify malicious intent behind digital interactions, rather than just detecting AI creation.
National Security and IoT Risks
Copied to clipboard!
(00:54:45)
  • Key Takeaway: The internet connectivity of critical infrastructure, such as solar inverters, creates millions of entry points that could be exploited by nation-states to cause widespread blackouts.
  • Summary: A deepfake of President Zelenskyy calling for surrender demonstrated the potential for large-scale disinformation in hybrid warfare. Research into IoT devices, like solar inverters connected to the internet, revealed vulnerabilities that could allow an attacker to seize control of massive amounts of electricity generation capacity. Connecting previously isolated grid components to the internet introduces national security risks via numerous entry points.
Hope and Reporting Crime
Copied to clipboard!
(00:59:12)
  • Key Takeaway: Despite the growing sophistication of cybercrime, there is hope as security firms and law enforcement cooperate to dismantle criminal rings and curb attacks.
  • Summary: The ongoing battle against cyber threats is a cat-and-mouse game where security technology will eventually catch up to malicious tactics, as seen with decades of malware defense. Victims are strongly advised to report scams, even if embarrassing, because underreporting (estimated at 93% for scams) prevents law enforcement from accurately budgeting and prioritizing the phenomenon.