A New Brunswick politician says he has learned a lesson after admitting he used artificial intelligence to help prepare a speech in the provincial legislature that turned into a viral video.
Bill Oliver, a Progressive Conservative, made the gaffe on June 9, as he read from speaking notes while delivering a speech that has garnered wide attention after it was shared in recent days on social media.
The speech includes a few odd references that turned out to be prompts from the AI tool he was using. This includes a moment in his speech when he paused and said: “Here’s a more natural, flowing version of that section that reads like a legislative speech rather than a series of short points.”
In a new statement released on Wednesday, Oliver recognized he hadn’t adequately reviewed the text before delivering his speech.
“When printing the final version of my speech, AI prompts were not removed, which were spoken by me and has caused much concern of many individuals,” he said in a statement released on Wednesday.
But Oliver’s blunder has prompted warnings about the erosion of public trust caused by the undisclosed use of AI.
While Oliver didn’t immediately catch his mistakes, the legislature’s recording of the entire incident reveals that no one else in the legislature appeared to notice or react to his use of AI prompts.
“AI has become a tool that is being used more often, in many areas, but should never replace our own ability to express our thoughts and convictions,” Oliver said in his statement.
“The sentiment of my speech was certainly mine, and I have learned an important lesson from this experience.”
The video spread around the world, generating headlines from Al Jazeera, Euronews, India Today and Kenya’s Tuko News, among others.
While it wasn’t included in the viral clip, Oliver later said, “here’s a more developed and flowing version of the section.”
Get daily National news
And, in an apparent third reference to AI, Oliver said, “this section is roughly,” — before cutting himself off and continuing to speak on the topic of affordability.
Jon Gillham, CEO of Collingwood, Ont.-based Originality AI, which uses software to detect AI for clients such as universities and news organizations, said he’s seen lots of examples of “AI residue” in text.
But, he said, Oliver’s speech was unusual.
“The confidence at which the AI residue was spoken in a public setting is unique,” he said in an interview.
“I’ve not seen that before. I think that’s certainly what helped make it go as viral as it did.”
Gillham said his company analyzed about 8,000 English language speeches delivered in the New Brunswick legislature between 2022 and 2026.
It found 3.3 per cent of speeches likely contained AI-generated text in 2022, rising to 21 per cent last year. So far in 2026, it found almost 23 per cent of speeches likely contained AI passages.
On at least one occasion, a politician asked a prepared question in the legislature, and received a prepared response back, with Originality’s analysis suggesting both were likely written with AI.
“So it was sort of political theatre with human actors reading AI narrative,” said Gillham.
Green Party representative Megan Mitton said Oliver’s speech affirmed her suspicions that legislators were reading AI-generated speeches throughout the first half of the year in what she described as a disservice to their constituents.
“We’re MLAs, we’re not elected to come here and read AI-generated speeches,” she said by phone Wednesday.
“We’re elected to be the voice of the people we represent, not the voice of AI.”
Mitten said the Green caucus has a rule against AI speech writing and would support a ban for all lawmakers.
The Liberal caucus said it does not have an AI policy but its representatives are “responsible for complying with their existing obligations regarding ethics and professional conduct.”
Mitten said the legislature needs “a higher standard” and could show leadership by creating a policy for the responsible use of artificial intelligence.
“When you look at public opinion polls, there’s already a crisis in our democracy. And I think that using generative AI instead of our own thoughts further adds to that.”
Gillham said the AI model politicians use matters.
Grok, for instance, is controlled by former Trump administration official and billionaire Elon Musk, while models like OpenAI and ChatGPT may have different political leanings.
“Or is it a Chinese open model, which is controlled by a political party?” Gillham said.
“I think all of those lead to the potential for manipulation, and so I think best practice is disclosure.”
Sarah Laframboise, executive director of non-profit Evidence for Democracy which tracks science and technology use in government, said politicians using AI for speeches has become common.
“The lack of human oversight is incredibly problematic, especially when we’re talking about things that are going into the legislative agenda and going into public records,” she said in an interview.
Laframboise’s organization in March released a report on AI’s use in policy making. It found that transparency was “shockingly low.”
She said the federal government is doing better than the provinces, with an online registry showing every use of AI in the public sector, including government vendors.
However, outside of election campaigns, Laframboise said there are very few AI guardrails for politicians.
She added there’s a code of conduct for federal MPs to which an AI policy could be added, but enforcement would be a major issue.
For instance, some phone cameras use AI in their image processing, so politicians may not even be aware they’re using the technology.
“I don’t know if we have a clear understanding of how to regulate that yet.”
“The tip-off was the same as all of us seeing what we have seen in our own conversations with ChatGPT or other AI LLMs”
I honestly wouldn’t know. I have no use for this nonsense.