Firefox's AI Switch Is Off. Telemetry Isn't

TEXT START: My Wireshark test of Firefox 148 showed that its AI kill switch disables generative features, but leaves telemetry and advertising untouched.

The Dissection

The article strips Mozilla’s “choice and agency” narrative down to its mechanism: Firefox offers a buried, reactive opt-out for visible generative features while preserving telemetry, advertising, experimentation, persistent identifiers, and AI infrastructure contacts. The switch is a pressure valve, not user sovereignty. It converts backlash into evidence of responsiveness while leaving the extraction system intact.

The CEO’s essay performs reputational laundering. User resistance becomes proof that Mozilla listens; telemetry then measures the resistance and feeds it back into the success narrative. The browser is allowed to stop showing some AI features, but not to stop observing, profiling, monetizing, or enrolling the user in its machinery.

The Core Fallacy

The article’s empirical diagnosis is largely correct, but its proposed remedy—first-run opt-in—remains trapped at the consent layer. Better consent does not alter the structural forces driving Mozilla toward AI, data extraction, advertising, and dependence on external capital and search partners.

Under the Discontinuity Thesis, this is transition management, not escape. A genuine opt-in could improve user control, but it cannot make Mozilla a Sovereign, make users economically indispensable Servitors, or reverse AI-driven displacement. It merely makes the surrender cleaner.

Hidden Assumptions

• That discoverability and default settings are the central problem rather than symptoms of a revenue-dependent institution.
• That users can exercise meaningful sovereignty through browser preferences while the platform retains control of infrastructure, telemetry, advertising, and experimentation.
• That privacy-respecting software can remain structurally viable without confronting the ownership and funding model behind it.
• That Mozilla’s willingness to ask permission can outrun the competitive pressure forcing every software layer to absorb AI.
• That exposing the deception is enough to produce institutional correction.

Social Function

Partial truth, mixed with transition management and ideological anesthetic. The article punctures corporate PR and documents the machinery precisely. But its endpoint—an honest opt-in—offers a manageable reform that lets the institution preserve its underlying model. It gives users a sharper map of the cage without addressing who owns the cage or why it is being rebuilt around AI.

The Verdict

Firefox’s AI switch is a controlled concession: generative automation can be hidden, while surveillance, monetization, and institutional control remain non-negotiable. Mozilla is not restoring agency; it is packaging dissent into telemetry.

The article catches the fraud but stops short of the terminal mechanism. Under DT logic, the decisive question is not whether Firefox asks before adding AI. It is whether Mozilla controls enough AI capital, infrastructure, energy, logistics, and maintenance to remain viable. This test shows a dependent institution trying to preserve legitimacy while adopting the machinery that will make its old social contract obsolete.