The impeachment trial of Vice President Sara Duterte has unexpectedly turned a spotlight on artificial intelligence (AI), particularly in relation to a video believed to threaten key government figures, including President Ferdinand Marcos Jr.
During the proceedings, several senator-judges questioned the authenticity of the video, with Senator Raffy Tulfo asking if it was AI-generated. Senior NBI agent John Mark Calilung responded that it was not, citing his experience and observation.
Calilung explained that the video's length—two hours—suggested it was not AI-generated, as longer videos typically require more complex technology that can lead to noticeable glitches. He argued that maintaining a convincing deepfake over such a duration is technically challenging.
Challenges of Verifying Digital Evidence
The trial has highlighted the difficulties in verifying video authenticity in an age where deepfakes are increasingly prevalent. Senator-judge Ping Lacson emphasized that merely relying on visual inspection to identify deepfakes is impractical.
Calilung acknowledged that while the NBI has access to advanced forensic tools like EnCase Forensic and Cellebrite, they opted not to use them for this particular video due to its length and the absence of obvious signs of manipulation during his manual review.
The absence of forensic analysis raises questions about the thoroughness of the investigation, especially given the current climate of skepticism surrounding digital content.
Expertise and Credibility in Question
While Calilung was deemed a competent witness, Senator Alan Peter Cayetano noted that he was not an expert witness on AI-generated content. The distinction is crucial, as verifying the authenticity of digital evidence often requires specialized knowledge.
The implications of this trial extend beyond Sara Duterte, as it reflects broader societal concerns about the reliability of digital evidence in legal contexts, particularly as AI technology continues to evolve.