What I cared about this week.
Canada’s National Artificial Intelligence Strategy: AI for All
Published 5 June 2026 (?)
The Canadian federal government’s AI strategy, penned by the Minister of AI Evan Solomon.
Thoughts
I have been skeptical about the Minister of Artificial Intelligence. I am not convinced that many of the core terms used in discussions surrounding AI are well-defined enough to support any real regulatory work. This report justifies my lack of confidence.
The report reads: “To use [AI], [Canadians] need to trust it.” I disagree on premise. Trust can only exist when both parties can be held accountable. We cannot hold AI accountable, nor can we hold anyone in the pipeline accountable. There is no potential for trust.
This report uses many words and terms that are ill-defined in an attempt to peddle off AI as some sort of magic solution. From the Minister of AI, I’d really expect more nuance. Though my expectations are inappropriate—most of the proponents of AI I’ve spoken with seem enthralled by the technology. They are not visionary, and cannot see beyond their nose. The issue I have then is my expectation, which I can adjust.
Also, as typical with Government of Canada reports, it is not entirely clear when this was actually published. It was last modified on 8 June, though the public PDF was created on 5 June.
Meta Silently Added Face-Recognition Code for Its Smart Glasses to Millions of Phones
Published 4 June 2026
Meta accidentally shipped code for facial recognition in their smart glasses product. This facial recognition was designed to be a real-time tagging feature, in which anyone within the wearer’s field of view would have their face biometrics captured and stored locally.
Thoughts
First, let’s ignore the sensationalistic tone of this report. Certainly, it’s important to call out when companies silently do things, but especially when it comes to code, I don’t think it’s all that difficult to accidentally add or remove things depending on workflows and pipelines. This story poses a question that supports my understanding of privacy.
I define privacy as the ability to say no. If you cannot say no, then you cannot have privacy. I like this simple framework because it removes needless theoretics and mental gymnastics, and pulls privacy back down to earth. It also supports legal analysis—it’s not a question of “would a reasonable person have privacy here”, but instead “would a reasonable person be able to say no”.
And now, the quote from the article:
If activated, it will transform faces captured by Meta’s glasses into unique biometric signatures, commonly known as faceprints, and check each one against faceprints stored on the user’s phone—a database that’s currently configured to receive updates from Meta.
There is a lot going on here! So let’s do a bit of a material assessment:
There are two main actors here:
The individual wearing the glasses
The glasses
There are two subjects here:
The public
The individual within a public space
An individual wears the Meta glasses and goes into the public. By virtue of their decision to wear these glasses with the configuration to tag people, the individuals within public space are having their faces scanned and translated into a faceprint—a biometric signifier.
Individuals in public spaces do not leave their homes assuming that their bodies will be transformed into manipulable data by other individuals in public. Can an individual say no to their faces being transformed into biometric data? Are there circumstances where individuals can say no?
I think it’s significant to change the tone of this article to what I set out above. This article, for example, jumps straight to who may use a biometric faceprint, and remains silent on what they may do—inviting the reader to come to their own conclusions. But unfortunately, this gets to the problem too late. I don’t care about what people can do with faceprints (especially illegally—I have little control over how people decide to break the law…). I care primarily about the creation of said faceprints. How they’re used is the next problem (and one I’d like to avoid from the outset).
The problem with this technology is not a faceprint, or how “invasive” it is (notably contestable, as a proponent of this tech will bring up the ‘how is your face private’ argument, which has legs). The problem is I have no way to say no.
This generates a healthier discussion—why would you want to say no? Well, now we can discuss the reality of data, how insecure many systems are, and how technology like this represents more of a slippery slope where convenience is selected over rationality. You take away a lot of my agency in public just so you can avoid remembering people’s names and faces.
So: sensationalistic reporting, but another brick in my house of “privacy is about saying no”—right beside webscraping and licenses plate cameras.
BILL C-34: An Act to enact the Digital Safety Act and the Digital Safety Commission of Canada Act and to make consequential amendments to other Acts
Published 10 June 2026
Canada’s Digital Safety Act, which is an attempt to regulate social media services and set up a digital regulator.
Thoughts
I am writing an essay on this, tentatively called the sex problem, in which I critically examine the role of sexuality in regulatory discussions. I draw the lines between reproduction as a human activity and reproduction as a state activity, lightly drawing upon Deleuze and other thinkers.
I will link it here when it’s done and posted.
Xiaomi’s new open source, agentic AI coding harness MiMo Code beats Claude Code at ultra-long, 200+ step tasks
Published 11 June 2026
Chinese company Xiaomi reports that its model MiMo beats out Claude in a particular benchmark activity. This claim has not been independently tested. Xiaomi’s MiMo is an open source model.
Thoughts
Interesting because Xiaomi implemented what I had sort of gestured to in my essay on AI as an extension of thought by prioritizing note-taking. I wonder if we’ll see more developments with peripheral solutions to create a form of memory that seems more like a database.
Also interesting because in some of my prior writing, I had said I feel as though I am a database with a faulty API. Of course, there is no reason to trust any company’s self reporting on product performance, so grain of salt.
This is a command-line agent, but it isn’t a local model. Users can opt to give voice commands, but only when logged in. Obviously, I’m curious about data privacy and its security.
As a final thought: a 200 step task seems a little bit ridiculous. I cannot imagine a task where I pass of 200 steps to a model without wanting a constant feedback. AI is a tool with a mouth—the benefit is the language-based feedback.
Statement on the US government directive to suspend access to Fable 5 and Mythos 5
Published 12 June 2026
Anthropic disabled access to Fable 5, its updated model, and Mythos 5, due to the U.S. government penning a directive which imposed export controls. The directive forbade the use of both models by foreign nationals. To keep in compliance, Anthropic disabled the models for all users.
Thoughts
This is American imperialism at its finest, signalled by the utterly fascinating double-think at play. There is a lot of noise around this situation, and all of it seems to revolve around personalities at the White House and Anthropic. Though it’s important to cut the ego-fat away and look at the bones and muscles.
First, the U.S. government’s current administration did not order an outright ban, but only for foreign nationals. There are a couple things to note here: (i) there is a capacity to single out a user’s status as American citizen or not—which has some sub-notes to discussion—and (ii) the administration was fine with allowing Fable/Mythos to persist so long as only American citizens could use it, which is rife with a lot of implication.
Second, there is a big problem with availability and sovereignty. I believe that sovereignty cedes away into the space of myth as computers forcefully pull the world together and stitch together jurisdictions without any potential for meaningful distinction. Yet still, we see a movement towards the strengthening of distinction, likely to the detriment of everyone. That the U.S. government had the ability to block access to a model to all peoples foreign to its realm of immediate control should form a fundamental concern about how AI is being developed and supported from the infrastructural perspective.
The foreign nationals piece
(i) I have a couple of data-related questions that stem from this story, as there is the strong implication that blocking foreign nationals is possible. This would suggest that Anthropic has enough data to determine whether the user holds American citizenship. It isn’t like this is super surprising from a data perspective, but the enriching of, for example, a user list against a U.S. citizen list suggests the bureaucratic process to do so. Or, more concerning, it suggests that such a capacity could be an executive order if this has no clear place in existing legislation.
I’m not a lawyer, nor do I really care about the law, but that this was an order suggests its possibility. It is not difficult to see how the shape of this order could form in other types of activities and outcomes. This, to me, underlines the importance of data security and data minimization. All of the sudden, what was once a “we don’t need need this, but it sort of helps, and why not” becomes the loophole for reams of unanticipated activities.
(ii) That the order was not an outright ban, and only for foreign nationals, suggests that the administration was not all that concerned with the potential damage that Fable’s “jailbreak” could cause. I find this point particularly interesting, as supposedly, U.S. officials brought up Dario Amodei’s comparison of Anthropic to a nuclear bomb. In spite of this, there was clearly no concern that a U.S. citizen would use the model to perpetrate anything nasty (I guess we’re just sweeping away domestic terrorism now, convenient). Evidently, the tone from the U.S. government was hot air—there was absolutely no concern for safety and well-being, otherwise the order would have been to remove access for everyone.
One could certainly read deeper into why the U.S. government’s order came as it did, and there are a couple pathways which point to plausible concerns. I’m ending this here though.
The sovereignty piece
I had kicked around the idea of doing a PhD, and one of the topics I circled briefly was AI sovereignty. This came from a burst of energy I experienced as I first used Claude—matched by a concern that this tool could be taken away from me, be it by my state or some other institution I have little to no control over. Months later, I read about that exact concern rearing its head a little bit.
I am increasingly interested in local models, as they seem (bolded seem) to provide more independence and reliability. I think everyone should see what just happened as a real concern going forward. States will always see technologies as theirs to muck about with. That technology has such a strong, pronounced role in the zeitgeist of economics and daily life exacerbates this.
The appropriate representatives from other jurisdictions have chimed in, calling for sovereignty. But as I just remarked, the state does not have your best interest in mind (neither do corporations, or any political structure for that matter). I read this situation as a need to develop my own resilience in the face of these powers acting as they see fit.
General thoughts: online safety
Thoughts
The news cycle had a lot of opinions on what online safety, online harm, and children safety online looks like. I struggle to keep my head cool, as I see a lot of assumption of experience as opposed to lived experience.
I am more of an anarchist than any other label. This is because I know my experience is not sufficient to comprehensively understand other people, so I would not be able to make decisions for others with any integrity. I have not met anyone who has some sort of complex understanding of other human beings that achieves this integrity.
I am reasonable—I know that something has to be done—but I am at my limit of people speaking as though they have the right solution for a nuanced situation where they have limited lived experience. As opposed to a more restorative approach, I see a lot of paranoid “problem solvers” put their make-believe hats on, and the cost is freedom and, by and large, queer identity.
Friction is necessary for growth. This much is obvious, and has been a fact throughout history. What is a coming of age novel, for example, if not an account of the friction necessary for one to grow? The well-intentioned but ill-conceived notion that we must protect children from the internet (which, in some ways, is a potent example of the real world) is going to rot the capacity of this next generation—if not push the individualistic children to more dangerous places.
But, I suppose this has been the playbook for… the last 20 years? More? Maybe always. Anyways, I read the impulse to clamp down on children relatively fascist, which is not surprising given how culture has increasingly softened to fascism now that we have enough distance between us and the last fascist catastrophe that meaningfully touched everyone (obviously, the west hasn’t been holistically embroiled in conflict).