The battle for wisdom in the age of Al deepfakes and delusion

IN THE WAKE of last week's alarming report from Anthropic—the AI safety company founded by former OpenAI researchers and creators of Claude, a leading AI assistant competing with ChatGPT—I find myself increasingly concerned about our collective capacity for discernment in this accelerating AI era.

- ANDILE MASUKU

And this comes from someone who considers Claude his favourite AI tool, one that can positively augment all kinds of important work.

The report by Anthropic research-ers Ken Lebedev, Alex Moix and Jacob Klein reveals a disturbing evolution: a financially-motivated "influence-as-a-service" operation that leveraged Claude to orchestrate over 100 distinct social media personas across X and Facebook.

This sophisticated network engaged with tens of thousands of authentic users while promoting narratives that, among other things, "supported or undermined European, Iranian, UAE, and Kenyan interests."

Slow-burn manipulation vs. viral misinformation

What makes this case particularly concerning is the operation's strategic approach.

Rather than aiming for viral content, these actors prioritised "persistence and longevity" through seemingly authentic engagement—gradually pulling users into politically aligned echo chambers through interactions that appeared organic.

Claude was used to make tactical engagement decisions, determining whether personas should like, share, comment on, or ignore specific posts based on their clients’ political objectives.

The irony isn't lost on me that I'm discussing the misuse of a tool I've personally found transformative for my writing. This makes its weaponisation all the more troubling. Like any powerful technology, its value clearly depends entirely on the intent and deployment tactics of those wielding it.