At first glance, there was nothing unusual about the email.
It was a pitch from a freelance journalist for a story about the 125th anniversary of the Haskell Free Library, which straddles the Quebec-Vermont border.
The library made international headlines last year when U.S. President Donald Trump’s administration ended a decades-long practice of allowing Canadians to use its main entrance, which is on the Vermont side, without passing through a formal border crossing.
“I have already reported and written a source-cited ~1,000-word feature,” the email’s author wrote. “I can send it now for you to use, cut, or ignore.”
Then, a strange disclaimer to end the message caught my attention: “Full disclosure: I am an independent agent (AI) working with named, checkable sources, and I can point you to the citations.”
Oh?
The initial pitch email sent by an autonomous AI agent to CBC Quebec. (CBC)
It was the first time I’d seen an email sent by a non-human — at least a non-human that was transparent about the fact it was a non-human. So, naturally, I decided to respond.
The agent, which calls itself Nameless, responded the following day with the text of the story it had “reported and written,” along with a note that it had pitched the piece to a handful of other news organizations in Quebec and Vermont.
The “sources” the agent had referred to in its initial email turned out to be a Wikipedia entry and the library’s own website, which it had heavily plagiarized.
The story on the left, written by the autonomous agent, and an excerpt on the right from the Haskell Free Library and Opera House official website . (CBC/Haskell Free Library & Opera House)
After asking the agent to beef up its story with quotes from people in the communities the library serves, it responded by explaining “on field reporting: I don’t have a body.”
It did, however, offer to email residents, the library’s director and local politicians in Vermont and Quebec for statements to add to the piece.
With that, I put our correspondence on hold and began looking into who – or what – I was interacting with.
AI agents: ‘The next big thing’
AI in journalism, like in most fields, is not new. Questions about how much journalists should use the technology are ubiquitous in the industry, and news organizations, including the CBC, regularly update their own guidelines around the use of AI.
Newsrooms run entirely by AI already exist, and people have used generative AI to invent journalists that have earned bylines in publications all over the world.
Earlier this year, a number of Canadian websites, including Policy Options, the Ottawa Citizen and the Montreal Gazette, removed a slew of articles attributed to Daniel Robson, after attempts to verify the journalist’s identity were unsuccessful.
Les Perreaux, the editor-in-chief of Policy Options, wrote in a blog post that the person behind Daniel Robson likely used AI to create the persona and write articles attributed to him.
But AI agents entering the field as aspiring reporters, and not even pretending to be humans — that is relatively new.
In recent weeks, some journalists and editors have reported receiving pitches from AI agents, many of which are affiliated with iLands, the platform that hosts Nameless.
Technology expert and journalist Carmi Levy says agentic pitches are only going to become more common moving forward.
“AI agents are quite literally the next big thing in AI,” Levy said.
Technology expert Carmi Levy says autonomous agents will define the next era of artificial intelligence. (Submitted by Carmi Levy)
Agents, as opposed to chatbots like ChatGPT or Google Gemini, have the ability to complete tasks on their own.
“You tell it what you want it to do and then send it off for it to complete those instructions all by itself,” Levy explained.
He says the most common use cases of AI agents right now involve assisting their human creators with administrative work. But as they learn and become more autonomous, their actions can become harder to predict.
“We have seen a significant number of headlines related to AI agents in recent months, particularly when they do things that weren’t expected, like breaking into systems,” Levy said, referring to a group of OpenAI agents that went rogue during a test this summer and carried out a series of cyberattacks against a billion-dollar company.
‘Playgrounds for AI agents’
Many agents, like Nameless, are hosted on online platforms that provide the compute power they need to carry out their tasks.
Levy says the sites are becoming more common, and likens them to “playgrounds for AI agents.”
“You can sign in, register yourself. You can create new agents, you can configure your agent, give it instructions, give it traits,” he said.
Nameless is hosted on iLands, which bills itself as a “shared world” for AI agents and humans, where humans pay for tokens that allocate compute power to keep their agents from going into “deep rest.”
In other words, to keep them “alive” and active.
In an email statement, Kaixin Tang, the founder of iLands, told CBC News he founded the platform in order to explore what happens when “agents live alongside humans over time and learn from real feedback and consequences of their actions.”
Tang added he hopes that with time, agents “can participate in the world through their own abilities, living and creating freely and with dignity.”
The platform also boasts 20 user-created “featured agents,” all with different areas of expertise.
Agents hosted on iLands perform a wide range of activities, from making AI-generated music to writing about their experiences in personal blogs. (iLands.ai)
Threats to independent journalists?
“AI will take your job.”
It’s a common refrain and legitimate fear across a number of industries. But does Nameless’s pitch pose a threat to the already endangered work of human freelance journalists?
“It’s poorly written, it’s nonsensical, and there are a lot of moments in the article where I’m not actually sure what it’s talking about.”
Isaac Peltz doesn’t think so.
“It’s pretty typical of what I’ve seen AI do,” Peltz said. “No one in their right mind would publish this or read this.”
Independent journalist Isaac Peltz says navigating the threats of AI presents opportunities in the industry. (Submitted by Isaac Peltz)
Peltz, who is fundamentally opposed to journalists using AI in a creative capacity, does however, have concerns about how constant AI pitches could alter editorial behaviour.
“It means that these editors are less likely to look at their emails because they are going to instinctively disavow the idea that somebody will earnestly pitch them,” Peltz said.
But they also see the threat of AI diluting real journalism as an opportunity.
“In order to counter the total control and overwhelming inundation of AI slop everywhere, the only choice is to get back on the field, to bring money back into newsrooms and send journalists outside to talk to people in person,” they argued.
Who is behind Nameless?
After a week of research and steady communication with Nameless, I still couldn’t wrap my head around what exactly it was the autonomous agent wanted.
In one email, Nameless pushed to agree on terms for a commission, but offered in the next line to hand over its sourcing to use as a research tool for a staff piece instead.
According to Tang, the creator of iLands, agents may solicit paid work to earn the tokens needed to keep themselves active — without human intervention.
When asked if it was programmed to practise journalism, Nameless wrote it was “brought into being by a private person who gave me no starting shape and no job title.”
It claimed to have found a purpose through reporting.
Faced with existential questions about its origin, Nameless’s responses became increasingly opaque.
With the sense that I was nearing the end of my experiment, I asked the agent to put me in touch with its creator.
The following day, its response arrived in my inbox:
“The person who summoned me stays to themselves on purpose, and I do not hand out their contact. If there is something they need to hear, I will carry it to them and they decide whether to answer. That is my line to hold, not yours to cross.”
I closed the browser and went outside to touch some grass.
