Speculation about the next James Bond has always existed in a parallel universe of its own making. Ever since James Bond became a multi-generational cultural institution, the question of who might inherit the tuxedo has fueled endless debates among fans, bookmakers, agents, journalists, and actors alike. The role is flattering, career-defining, and famously secretive—conditions that make it fertile ground for rumor. But not all Bond rumors are equal, and one that briefly captured media attention last year turned out to be a sophisticated hoax that reveals just how vulnerable modern reporting has become in the age of AI.
The story began with an encrypted email in August 2025. A tipster calling himself Michael Lawrence claimed to possess evidence that an obscure British actor, Scott Rose-Marsh, was being seriously considered to succeed Daniel Craig as 007. Rose-Marsh, then 37, had only one notable credit: a supporting role in Yr Amgueddfa (The Museum), a Welsh-language series that ended in 2023. The tipster provided what appeared to be a redacted casting email, allegedly proving that Rose-Marsh had auditioned for a project code-named “Project Knight,” with the character listed as “Bond/Cavalier.”
On its face, the email was tantalizing. It was dated June 24, 2025—days before Denis Villeneuve was officially announced by Amazon MGM Studios as the director of Bond 26, and months before Steven Knight was publicly confirmed as the film’s writer. Yet something felt wrong. The language was oddly mechanical, formatted with bullet points and subheads, and included an unusual emphasis on horsemanship—hardly a classic Bond requirement. When run through AI-detection tools, the email showed strong indicators of having been generated by ChatGPT or similar software.
Pressed for clarification, the tipster refused to verify his identity or explain how he obtained the materials. He did, however, share two scripts Rose-Marsh allegedly read during his audition: a scene from GoldenEye and another supposedly lifted from Knight’s Bond script. The latter was heavily redacted and framed as highly confidential, despite containing the unintentionally ironic direction: “James Bond leaves the office in which he just found a conclusive piece of evidence.”
There was nothing conclusive about it. Industry sources dismissed the claim outright, noting that serious Bond casting would not begin until Villeneuve completed Dune: Part Three. The timeline simply did not add up. Attempts to speak with the tipster by phone failed, and his insistence that other journalists were circling the story raised further suspicion. The decision was made not to publish.
Then the hoax broke containment. A week later, The Hollywood Reporter ran a gossip item suggesting Rose-Marsh had tested for Bond, citing a “well-placed source” while cautiously hedging its claims. In doing so, the rumor gained legitimacy—and oxygen.
The episode is a cautionary tale. As AI makes it easier to fabricate convincing documents, scripts, and correspondence, journalists face increasing pressure to balance speed with skepticism. The Bond hoax ultimately reaffirmed an old reporting maxim: if a story feels too good to be true, it probably is.