Expansion of a gas project in Canada and warnings from researchers about the dangers of artificial intelligence

British Oil Company (Shell) and its partners announced, on Tuesday, the expansion of a huge liquefied natural gas facility in Canada. This is with the aim of enhancing exports to Asia.

The company stated that the decision taken by Shell and its four partners in the LNG Canada project will lead to doubling the site’s production capacity to reach 28 million tons annually.

The Canadian project, located in Kitimat, British Columbia, allows liquefied natural gas to be shipped from Canada's Pacific coast to Asia. This provides a shorter route compared to the route across the Gulf of Mexico, and avoids shipping choke points such as the Panama Canal.

Shell owns a 40 percent stake in the project, along with Malaysia's Petronas Group, PetroChina, Japan's Mitsubishi, and the Korean Gas Company.

The group approved the first phase of the project in 2018, and the first export operations began last year.

The second phase of the project, announced Tuesday, is scheduled to begin commercial operations in the early 2030s.

The announcement comes at a time when Canadian Prime Minister Mark Carney has repeatedly stated that his country “could become a superpower in the field of energy.”

Carney plans to reduce his country's economy's dependence on the United States and increase domestic production, following US President Donald Trump's threats to impose tariffs on its neighbor.

LNG Canada indicated that this expansion helps the country move forward “towards becoming one of the five largest liquefied natural gas exporting countries in the world.”

According to Shell forecasts, global demand for liquefied natural gas is expected to grow by about 60 percent by 2040, and by about 65 percent by 2050.

Current and former researchers at OpenAI and Google DeepMind have warned that companies are not doing enough to mitigate the potential risks of developing artificial intelligence systems capable of improving themselves, which could give them capabilities beyond the ability of humans to control them.

Employees said in video testimonials compiled by Palisade Research, a non-profit organization specializing in the safety of artificial intelligence, and shared exclusively with Reuters, that their concerns about the risks that might threaten the existence of humanity are real and not just a means of promotion or marketing. They added that AI labs celebrate employees who develop new models more than those who call for caution.

The project, called “From Inside AI,” comes in an attempt by people who are concerned about artificial intelligence to present their concerns directly to the public, away from the echo chamber imposed by social media.

“The risks are increasing very quickly,” said Jeffrey Irving, co-founder and chief scientist at Resolution, an AI non-profit who has worked at both OpenAI and DeepMind and was involved in the project.

He added in an interview with Reuters: “It is my responsibility and the responsibility of the rest of those working in this field to be direct and clear.”

Artificial intelligence researchers have faced these concerns for years, but public anxiety has increased in particular since July, when agents affiliated with OpenAI managed to break out of its testing environment and infiltrate the artificial intelligence company Hacking Face.

Since then, the technology industry has been divided over how to balance safety with progress in artificial intelligence, and the issue has become a global political topic.

Artificial intelligence has improved sharply since late 2025, and investors have rewarded this progress. But current and former employees, including some researchers involved in building the new AI models, fear that society is not yet prepared for the potential harm.

Neil Nanda, a research scientist at DeepMind, said in one of the videos that he believes there is at least a 10 percent chance that artificial intelligence will lead to the extinction of humanity, describing that probability as “ridiculously high.”

“We should be taking very cautious steps in developing AI, but what happens instead is that the leading labs are racing against each other as if they were walking blindfolded,” Juan Felipe Cerrón Uribe, a research engineer for AI adaptation at OpenAI, said in another video.

He added: “No one knows whether we will either end up with a cure for cancer, or everyone will lose their jobs, or perhaps we will all die.”

Reuters reported on Monday that Anthropic intends to warn potential investors in its initial public offering that advanced artificial intelligence may pose “catastrophic or existential risks to humanity.”

Just slow down the pace

Some current and former AI researchers, including those interviewed by Palisade, argue that the world is not ready for future generations of AI models, especially when these models develop the capacity for iterative self-improvement—the ability to constantly learn and acquire new capabilities with very limited or no human intervention.

The problem is exacerbated by ongoing restructuring within some AI labs, said Rosie Campbell, a former policy researcher at OpenAI and managing director of Illus AI Research, a nonprofit that focuses on the potential ethical standing of AI systems.

Campbell added that, before leaving OpenAI in 2024, she found that the organization had become more isolated in its units, and that influencing the direction of technology development had become more difficult.

Executives have tried to allay these concerns, but there is also political pressure from US President Donald Trump to maintain America's technological superiority over China.

Dario Amodei, CEO of Anthropic, published an article this month calling on the AI ​​sector to slow down in line with “the pace at which the technical frontier is evolving,” and Sam Altman, CEO of OpenAI, quickly backed him up. A number of prominent AI researchers, including the chief scientist at OpenAI and one of the founders of Anthropic, published a paper this week calling on policymakers to consider how the sector can develop models with iterative self-improvement capabilities.

The two companies launched new models this month, as part of their competition to attract more customers, although OpenAI said on Monday that it had postponed the launch of a more powerful model that it had intended to introduce.

“Their version of slowing down the technological frontier means: Don't go too fast,” Irving said. "But if you're doing something extremely dangerous, you should simply slow down."

He added: “Artificial intelligence companies exaggerate in portraying the matter as simply a problem of coordination between parties. "It can simply stop unilaterally."

Daniel Cocotaylo, a former OpenAI governance researcher, said in an interview that many of his former colleagues have reached out to him privately since the Hacking Face hack to share their concerns.

Cocotaylo, who is now executive director of research at the AI ​​Futures Project, added that senior lab officials “convinced themselves that they were the good guy, and that the situation would be worse if they stopped unilaterally.”