Fake Minns, altered images and psyop theories: Bondi attack misinformation shows AI’s power to confuse | Bondi beach terror attack
In the hours and days following the terror attack on Bondi beach, it was difficult to avoid misinformation powered by artificial intelligence, with some platforms making dubious claims to users trying to find factual information.
The “for you” page that those behind the attack were IDF soldiers; those injured were crisis actors; that an innocent person was one of the alleged attackers; It was stated that the Syrian Muslim hero who fought the attackers was a Christian named English.
Generative AI has made the situation even worse.
An altered clip featuring deepfake audio of New South Wales premier Chris Minns making false claims about the attackers was shared across multiple accounts.
In one particularly egregious example, an AI-generated image based on a real photograph of the victims was altered to suggest that she was a crisis actor with red makeup applied to her face to look like blood.
Arsen Ostrovsky, the human rights lawyer depicted in the fake photo, later said on
Pakistani Information Minister Attaullah Tarar said his country was the victim of a coordinated online disinformation campaign following the bondi beach terror attack and false claims were circulating that one of the suspects was a Pakistani national.
The falsely identified man told Guardian Australia it was “extremely disturbing” and traumatic when his photo spread on social media, along with allegations that he was the alleged attacker.
Tarar said the Pakistani man was “the victim of a malicious and organized campaign” and claimed that the disinformation campaign originated in India.
Meanwhile, X’s AI chatbot Grok told users that the hero who captured and disarmed one of the alleged gunmen was an IT worker with an English name, rather than the Syrian-born Ahmed al-Ahmed. It appears that this claim originates from a website established on the same day as the terrorist attack to imitate a legitimate news site.
Images of Ahmed created by artificial intelligence also spread rapidly on social media. crypto schemes and fake fundraisers.
It was a far cry from Twitter’s heyday as the hub for breaking news. Misinformation was around back then, too, but it was less widespread and wasn’t served through an algorithm designed to reward anger-based interaction (especially for verified accounts that would benefit financially from that interaction).
Many of the posts featuring false claims have been viewed hundreds of thousands or even millions of times.
Legitimate news was circulating in X, but it was buried under AI-powered misinformation.
When Musk took over X, he scrapped the site’s fact-checking plan in favor of a user rating system called “community ratings” that added crowdsourced user verification to posts. Other platforms are following the same path. Meta removed the previous fact check system to implement its own version of community notes.
But as QUT lecturer Timothy Graham said this week, the community grades system is not useful where opinions are deeply divided. It takes too long. Community notes have since been applied to many of the examples above, but this happened long after most people had seen the original posts in their feeds.
X is trying to get Grok to create its own community notes to fact-check posts, but if we’re to take Ahmed’s example into account, this is even more concerning.
The company did not respond to questions about what it is doing to combat misinformation posted on its platform or spread by its AI chatbot.
One saving grace is that most fakes are easy to spot for now. For example, the fake Minns’ accent had an American ring to it, making it clear that it wasn’t her. The crisis actor post had many hallmarks of dodgy AI rendering, such as misformed text on a T-shirt.
Media outlets mostly ignored these posts or called them out.
However, this may change as AI models improve, making it even more difficult to separate fact from fiction. Meanwhile, AI companies and the platforms that host their content seem complacent about doing anything to prevent this.
Digi, the industry group representing social media platforms in Australia, proposed removing the requirement to combat misinformation in industry rules earlier this year, saying “recent experience” showed “misinformation is a politically charged and controversial issue within the Australian community”.
It’s hard to see how this week will change things.
