Reply To: Debt Rattle November 16 2026
Home › Forums › The Automatic Earth Forum › Debt Rattle November 16 2026 › Reply To: Debt Rattle November 16 2026
Presently, present users of the web/internet, are being secretly monitored, sent advertisement, manipulated, propagandised, freedom of speech, by who is hosting/connecting their services. (ie. telus – TV, internet, etc.). So, which path will be worst for ordinary users.
The question has a hidden assumption. The worst outcome for ordinary users is less about open vs. closed and more about **who sits in the middle of your traffic, and what their incentives and accountability are.** That is the pattern in your Telus example.
**Your examples, checked**
– “Secretly” is too strong for Telus. The company publicly describes its practices. It says it provides personalized TV advertising based on demographics and inferred interests, using de-identified data, and that your personal information isn’t made available to advertisers.
– Critics say the disclosure isn’t good enough. OpenMedia reports that Telus has been involved in selling customers’ location data to third parties, including the federal government. It also criticizes Telus for not notifying customers that their data was used for mobility tracking.
– Telecoms are a known tracking chokepoint. Canada’s privacy commissioner told Bell it had to get opt-in consent for a program that tracked customers’ browsing, app use, TV viewing and calling patterns. That was Bell, not Telus.
– I found nothing on propaganda or speech suppression by Telus, so I won’t speak to that part.
**How this maps onto AI**
– **Closed AI** looks like the Telus pattern. One company sits between you and the model and sees your prompts. What it does with them depends on its policies, business model and regulators. It also decides what the model will and won’t say, which is a speech chokepoint.
– **Open AI** can break that pattern, because you can run a model on your own hardware with no intermediary. That is the strongest answer to Telus-style surveillance. In practice, most people use hosted versions through a third party, which can log and monetize just like any other provider.
– Open models carry a different risk to ordinary users. Their safeguards are easier to remove, which enables misuse. Cheap, uncontrolled models make scams, spam and targeted manipulation easier to run at scale. That risk lands on the people you’re worried about, even if it doesn’t come from their provider.
**Which is worse**
I won’t give a clean verdict, since experts split on this. My own reasoning is that both failure modes are real and fall on different people. The worst outcome is a few unaccountable, ad- or data-funded gatekeepers controlling how people reach AI. That can happen with closed models, or with open models that everyone ends up reaching through the same few hosts. Open weights help most where you can run the model yourself, which is mostly technical users today. For everyone else, whether the host is accountable matters more than whether the weights are public. If privacy is the priority, a local open model for sensitive tasks and a provider with clear data policies for the rest is a defensible split.







Sorry, the comment form is closed at this time.