AI-generated antisemitic deepfake images that are undetectable by software are gaining popularity, a study from Antisemitism Policy Trust found.
In a report titled "Detecting Deepfakes by the Antisemitism Policy Trust and Project Decoding Antisemitism," researchers found that “AI-generated antisemitism is on the rise, with potentially significant implications for the Jewish community.”
“Deepfakes” are artificially manipulated audiovisual material produced using an AI to create false videos or photos.
The Antisemitism Policy Trust compiled a gallery of 51 AI deepfakes with antisemitic images that are increasingly undetectable to software.
“Most existing classifiers exhibit a very weak performance when analyzing antisemitic deepfakes,” the report stated.
"Implicit patterns and ambiguities in an image that could lead to an antisemitic interpretation are not recognized because AI-based approaches do not take into account emerging practices of image use or manipulation when identifying deepfakes."
Researchers discovered that out of 51 deepfakes studied, 28 of them were classically antisemitic, depicting stereotypical Jews as evil characters. The rest demonized Israel for either emotional manipulation or disinformation purposes, the report claimed.
"Hiding hate symbols within seemingly innocent images has become a way to spread hate across the web, so much so that the practice arguably forms its own strand of generative AI design," the report stated.
For the latter, the report found that a frequent deepfake image garnered was "The depiction of blood-thirsty Israelis victimizing Palestinian children."
“AI tools are getting smarter, and yet those seeking to spread hate are still marching ahead of us, duping the technology and gaming the systems," said Danny Stone, head of Antisemitism Policy Trust, as reported by the Sunday Times.
“AI developers must put risk assessments and counter-hate strategies at the heart of their design," he continued. "Meanwhile, government needs urgently to show leadership, and fix our regulatory regime which is already being revealed as not fit for purpose.”
Unprecedented rates of antisemitism
The report comes amidst unprecedented rates of antisemitism worldwide, both online and in person.
The Jerusalem Post previously reported that the European Union Agency for Fundamental Rights (EUFRA) recorded an increase of up to 400% in antisemitic activity since the outbreak of the war.
Additionally, EUFRA also found that 96% of Jews encounter antisemitism in their daily lives.
Regarding online content, the EUFRA reported that the UK has the highest number of antisemitic posts, at around 2.5 million, followed by Spain at 1.3 million and France at 1.15 million.