A public resignation from one of the world's leading AI labs pulled Ottawa's minister into the existential-risk debate on the same day he launched a $13-million literacy push in Edmonton.
Canada’s AI minister, Evan Solomon, said on Wednesday that the federal government is watching developments at Anthropic closely after a researcher resigned from the company and publicly warned that people building advanced AI systems privately fear the technology could “kill us all.” Speaking to reporters in Edmonton on September 9, 2026, Solomon said the warnings were not new and that both regulatory and technical remedies are available.
The exchange put an internal dispute inside a San Francisco AI lab directly onto the Canadian political agenda — and it happened on the same day the Minister of Artificial Intelligence and Digital Innovation was in Alberta announcing a national push to raise everyday AI literacy.
The X Post That Put Ottawa on the Spot
On Tuesday, September 8, researcher Jacob Coxon — who has worked at both OpenAI and Anthropic — announced his departure in a post on X, writing that neither company “is acting responsibly” and that both are “racing straight to self-improving superintelligence and gambling with our lives.”
Coxon said many of the people building these systems “earnestly believe that it could kill us all by the end of the decade,” adding that the statement was “not a marketing stunt.” He argued that industry insiders tend to soften their language in public while voicing sharper fears privately.
The claim drew an unusual on-the-record confirmation from inside Anthropic. Evan Hubinger, a team lead at the company focused on AI safety, replied to the post: “we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade.”
That combination — a resignation letter in public view and a current safety lead attaching a double-digit probability to catastrophic outcomes — is what reporters brought to the minister the following day.
Solomon’s Four-Part Answer: Privacy Law, a Regulator, Safety Money and Allies
Solomon did not dismiss the concerns. He acknowledged there are “real concerns” about the negative consequences of AI, while framing the warnings as part of a longer-running debate rather than a sudden rupture.
His response pointed to four levers the federal government says it is pulling:
- Privacy legislation: Solomon cited the Liberal government’s work to introduce privacy legislation as part of the regulatory answer.
- A new regulator: “We’re establishing a new regulator that will have the power to hold these big companies to account,” he said. Further details on the regulator’s structure, timing or statutory powers were not provided at the news conference.
- Investment in AI safety: The minister said Ottawa is putting money into AI safety work, and specifically noted ongoing research by Canadian AI pioneer Yoshua Bengio on safe AI.
- International cooperation: Solomon said the government is working with other countries on approaches to protect citizens — an acknowledgement that no single national rulebook governs frontier model development.
Notably, the minister’s answer separated the problem into two categories: things a regulator can require of companies operating in Canada, and things that have to be solved at the level of the technology itself.
Why Yoshua Bengio’s Name Keeps Coming Up
General background for readers: Bengio is among the most-cited researchers in deep learning and one of the reasons Canada is treated as a serious jurisdiction in global AI discussions despite being a small player in compute and capital. When a federal minister reaches for a technical credential in response to an existential-risk question, Bengio’s safe-AI research is the most available Canadian answer.
That matters for how Canada positions itself. Ottawa cannot slow down a training run in California, but it can fund and host research on evaluation, oversight and alignment — and then argue for those methods in international forums. The source material does not specify funding amounts or program names attached to that safety investment.
The $13-Million AI Literacy Course Announced the Same Day
Solomon made his safety comments in Edmonton, where he was launching a new initiative to raise AI literacy among Canadians. The program includes “AI essentials” courses aimed at students, educators and the general public.
“We’ll invest $13 million to launch this course, to expand AI literacy for every single Canadian,” Solomon said. The federal government is delivering the initiative in partnership with the Alberta Machine Intelligence Institute, one of Canada’s three national AI institutes.
Readers following the rollout can find more detail in our earlier coverage of the national AI literacy initiative launched by Ottawa with Amii and the plan for a free AI course for Canadians before year’s end.
The juxtaposition was hard to miss: a minister fielding questions about superintelligence risk while announcing an introductory course on how AI tools actually work. Both sit inside the same portfolio, and both reflect a government trying to cover the near-term and long-term ends of the same file.
Voluntary Data Centre Principles — and Who Enforces Them
The safety exchange came a week after Solomon announced a set of voluntary principles for building data centres in Canada, released in response to growing public backlash against data centre construction across the country.
The framework says data centres must:
- Create lasting local benefits in the communities that host them.
- Avoid shifting electricity costs onto Canadians.
- Minimize water use and environmental impact.
- Provide transparency on local impacts.
- Bring strategic value to Canada, rather than serving purely as offshore capacity.
The federal government will not enforce the guidelines. Solomon said decisions on data centres fall under provincial and municipal jurisdiction, and that the agreement’s role is to provide reliable and verifiable information on the standards that were set. “Those are going to be enforced by the municipalities or the provinces,” he said.
The signatory list spans hyperscalers, Canadian telecoms and at least one frontier AI lab.
| Category | Signatories named |
|---|---|
| Global cloud and platform firms | Amazon, Google, Meta, Microsoft |
| Canadian telecommunications | Bell, Telus |
| AI model developers | Anthropic |
Anthropic’s presence on that list is part of why the resignation story landed as a Canadian policy question rather than a foreign business story. The same company named in the safety warning has signed on to Ottawa’s voluntary infrastructure standards.
How the Week Unfolded
| Date | Event |
|---|---|
| Week of Sept 1, 2026 | Solomon announces voluntary data centre principles; federal government confirms it will not enforce them |
| Tuesday, Sept 8, 2026 | Jacob Coxon announces his resignation on X, warning about the race to self-improving superintelligence |
| Tuesday, Sept 8, 2026 | Anthropic safety team lead Evan Hubinger publicly puts the risk above 10 per cent within the decade |
| Wednesday, Sept 9, 2026 | Solomon responds in Edmonton and launches the $13-million AI literacy initiative with Amii |
The Gap Between a Warning and a Statute
For Canadian readers, the substance of the week is the distance between the scale of the warning and the specificity of the response. A current safety lead at a leading lab assigned a greater-than-10-per-cent chance of catastrophic outcomes within a decade. The federal answer is a privacy bill, a regulator that has been described but not yet detailed publicly, safety research funding, international talks, a voluntary infrastructure framework enforced by other levels of government, and a $13-million literacy course.
None of those are trivial, and the literacy program in particular has a defined partner, a budget figure and a delivery route. But the enforcement question runs through all of it. Ottawa’s data centre framework explicitly relies on provinces and municipalities to act, and the promised regulator’s powers over “big companies” have not yet been spelled out in legislation.
The practical next milestone to watch is the privacy legislation Solomon referenced and the statutory design of the new regulator, since that is where the government’s stated ability to hold large AI firms to account would either become enforceable or remain aspirational. No introduction date for that legislation was given.
Frequently Asked Questions
Who is the minister responsible for AI in Canada?
Evan Solomon, the Minister of Artificial Intelligence and Digital Innovation, who made these comments in Edmonton on September 9, 2026.
What exactly did the Anthropic researcher say?
Jacob Coxon, who worked at both OpenAI and Anthropic, wrote on X that neither company is acting responsibly, that they are “racing straight to self-improving superintelligence,” and that many builders believe AI could kill everyone by the end of the decade.
Did anyone still at Anthropic confirm that view?
Evan Hubinger, an Anthropic team lead working on AI safety, replied publicly that he personally puts the risk above 10 per cent within the next decade.
Are the new data centre principles legally binding?
No. The federal government has said it will not enforce them; Solomon said data centre decisions fall to provinces and municipalities, and that the agreement’s function is to provide verifiable information on the agreed standards.
What does the $13 million actually fund?
It funds the launch of “AI essentials” courses for students, educators and the general public, delivered in partnership with the Alberta Machine Intelligence Institute. A detailed breakdown of the spending was not provided.
