Tips for teams to spot and protect against AI deepfakes

“Mom, I'm out to dinner with Jay, and we need to settle the bill. I can't find my phone; I think I left it in the car on the way here. Could you send him $80 for my share of the dinner? I’ll send you a payment link. Also, I still owe him $400 for the concert tickets. Could you add that in as well?” 

What? The FaceTime call catches me off guard. It's great to see my son, but the connection is fuzzy, and there's lots of background noise.  

“Thanks, Mom! I’ve got to go look for my phone. Love you!” The call ends abruptly, and I get a text from an unfamiliar number with a link to pay $480 for my son’s dinner and concert tickets. 

I pause and scrutinize the link, unsure of its legitimacy. While I, of course, want to help my son – something seems off here. What would you do? Would you send the money? 

What is an AI deepfake? 

Deepfakes are synthetic media, typically videos or audio recordings, created through artificial intelligence (AI). By altering or fabricating content, deepfakes can convincingly portray real people engaging in actions or making statements they never actually made. While this technology can serve benign purposes in entertainment or education, it has also been exploited with malicious intent. 

• Political manipulation: In 2018, researchers produced a deepfake video of Barack Obama, depicting him making statements he never uttered. This incident underscored the potential for deepfakes to sway political opinions and create discord. 
• Celebrity impersonation: Deepfakes have been used to generate non-consensual videos involving celebrities, raising ethical concerns and highlighting the risks to personal reputations. 
• Corporate espionage: Companies have suffered financial losses due to deepfake technology being used to impersonate executives, resulting in unauthorized transactions. In early 2024, a finance worker in Hong Kong transferred over $25 million to fraudsters who impersonated the company’s chief financial officer during a video call. 

The creation of deepfakes has become increasingly accessible due to technological advancements and the widespread availability of AI tools. Numerous online platforms and software applications offer user-friendly interfaces that require minimal technical expertise, enabling individuals without a background in AI to produce realistic deepfakes.  

Additionally, large datasets of images and videos, often freely available online, provide ample resources for training AI models to generate synthetic media. This ease of access poses significant challenges, lowering the barrier for creating misleading content. According to estimates from Statista’s Market Insights, the global cost of cybercrime is expected to surge over the next four years, rising from $9.22 trillion in 2024 to $13.82 trillion by 2028.

Can you identify a deepfake? 

Detecting AI deepfakes requires a discerning eye and knowledge of specific indicators. While deepfakes have become more advanced, there remain distinct clues that aid in their detection. It's crucial for individuals and organizations to become familiar with these signs, which include: 

• Inconsistent lip sync: Observe if the lip movements and spoken words are misaligned or appear unnatural. 
• Unnatural facial movements: Be vigilant for stiff facial expressions or movements that don't match the environment or context. Notice any irregular blinking or eye movements. 
• Visual degradation: Look out for flickering, blurring, or digital artifacts such as pixelation around the face or body, that suggest tampering. 
• Inconsistent lighting or dhadows: Deepfakes may show lighting mismatches that don't align with the rest of the scene. 
• Mechanical vocal tone: Listen for unusual voice modulation. Is the tone unnaturally consistent? Are there strange pauses or background noises? 
• Unverified source: Check the origin of the content. Was it shared by a reputable source? On which platform was it found, and what references support its authenticity? 

Ways to protect yourself from deepfakes 

Embracing skepticism and a critical mindset is essential for safeguarding against manipulated media. Cross-referencing information with reliable news sources or official organizations is crucial, and fostering a culture of investigative rigor within your team can ensure the credibility of information before dissemination. Organizations are advised to protect against AI-generated content by implementing a comprehensive program comprising the following elements: 

1. Education: Equip employees with the skills to recognize deepfakes and nurture a culture of skepticism towards dubious content. Train them to identify emotionally charged messages and scenarios that pressure quick responses. Awareness is the strongest defense against this kind of manipulation. 
2. Verification: Teach employees how to verify content efficiently using reliable sources. Utilize tools like Google's Fact Check Explorer and Snopes for this purpose. Encourage thorough investigations when inconsistencies arise by asking video call participants to turn their heads, use a code word, or re-establish communication via a known channel. 
3. Expert support: Set up a reporting channel where employees can voice concerns and seek support in cases of potential manipulation. Cybersecurity experts should be available to assist your team. 
4. Detection tools: Deploy AI-driven solutions such as Deepware Scanner and Sensity AI to analyze and detect possible deepfakes. 

By implementing these strategies, organizations can better protect themselves and their teams from the threats posed by AI-generated content. 

Should I send the money?

As I consider my child's request for assistance, two additional text messages arrive with a link for payment. Yet another sign that something is off. My mind fills with questions:  

• Why is a college student spending so much on dinner? 
• What concert is he referring to? 
• Is he safe? 

I attempted to call back the number but received no response. I then called his phone directly, and immediately my son answered. He’s in his dorm room, studying for tomorrow’s exam. The FaceTime call I received was a deepfake. After informing the family about the incident, I contacted the police. 

Protect yourself against deepfakes 

In an era where seeing is not always believing, awareness and vigilance are key. By staying informed about deepfake technology and its implications, individuals and business leaders can better protect their organizations and stakeholders. Remember, a proactive approach to digital security will help safeguard against the evolving threats posed by deepfakes. 

For a technical deep dive, learn how deepfakes are created here.