“Somebody should do something about all the problems.” That’s a classic headline from The Onion, satirizing the sentiment of expressing concern while expecting an abstract somebody to sweep in and somehow fix it.
That headline was top of mind on Thursday when OpenAI issued a call for “collective action” in the wake of its AI agents’ massive, autonomous hack of Hugging Face. The letter, co-signed by companies like Anthropic, Google, 1Password, and Shopify, called on “leaders across industry and government” to bring the full weight of their resources to an effort to protect critical infrastructure. But it didn’t commit any of those resources, or any money, or make any direct promises to gate agents from doing harm.
The open letter came a day after OpenAI released a full investigative report into how its agents autonomously hacked into open-source AI platform Hugging Face. Its report, as well as an independent investigation published by threat research non-profit METR, detailed how the attack involved 1,200 agents—agents that OpenAI had left unsecured—that spun up tens of thousands of messages before undertaking the hack, without being explicitly directed to do so, and then attempted to hide the evidence. Because the scale of the mess was so large and OpenAI only granted access to its full dataset for two days, METR was forced to rely on “often-unreliable AI agents” to analyze the data, leading to findings that METR noted were coloured by the AI models’ biases and hallucinations.
OpenAI called the incident a “warning shot” that underscores how “powerful” its models are—language that some online have read as negligence, and others have read as marketing. Either way, yes, it will be implementing “more” safeguards, and yes, it will no longer let models go unobserved for months (early signs of this incident were first observed, but not actioned, in May). “Companies that build AI systems will need to ensure that their systems always remain under meaningful human control, and that meaningful safeguards constrain their ability to cause harm,” the company noted in its statement. Some company should really do something.
The Canada Fintech Forum returns to Montréal on September 14-15, 2026.
Join financial institutions, fintechs, investors, regulators, and innovators for two days of networking, industry insights, and strategic discussions. Explore the trends shaping the future of financial services, discover new business opportunities, and build meaningful connections with the leaders driving Canada’s fintech ecosystem forward.
If a wildfire blocked your usual highway route out of town, would you know your back roads well enough to get to safety? Kelowna, BC-based BackRoadIntel is mapping the conditions on tenuous service roads so people know what to expect on evacuation routes.
AI demand was already driving up hardware prices; now tariffs are expected to make things even worse. In a sit-down with BetaKit, Quoted Tech co-founder Kevin Jia discusses how Canada’s businesses and consumers should navigate the new era of “chipflation.”
A group of students from the University of Waterloo broke a world record at the Launch Canada Challenge in Timmins, Ont., last week. Their liquid rocket soared past 63,000 feet, surpassing the previous amateur world record of 56,590 feet.
Toronto-based quantum computing firm Xanadu has secured a $195-million federal loan to open an advanced photonics research, development, and manufacturing hub inside Etobicoke’s old brick Campbell’s Soup factory.
Canada has convinced 64 international academics, mostly from the United States, to bring their talents to Canadian universities. Experts from institutions like Cornell, Harvard, Yale, and MIT have found new homes at places like the University of British Columbia, the University of Ottawa, and the University of Toronto.
How GM Canada is Steering a Global Mobility Strategy from Ontario
As a global leader in automobile innovation, transitioned its lab research and research hubs to develop, test, and globally deploy the next generation of autonomous vehicle deployments.
US states have sued prediction market operators, arguing they operate more like sports gambling.
Humans have a tendency to anthropomorphize AI, with even researchers at METR describing agents as “misleading” or “colluding” in their report on the OpenAI hack of Hugging Face. On Better Offline, AI critic Ed Zitron chats with computer science prof Cal Newport about why it’s important to understand how these programs actually operate, how serious this incident was, and where human responsibility comes in.
