Firefox Adds AI Off-Switch, Brave Charges $60 for No-AI, Argentina AI Stumbles

Crypto News

Mozilla revealed Project Nova on May 21, a sweeping visual overhaul of Firefox set to roll out later this year, but the headline feature isn't the rounded tabs or the fire-inspired color palette. Buried in the new settings panel sits a single, plain-language toggle that disables every AI feature in the browser at once — no hunting through nested menus, no dark patterns. The redesign reintroduces compact mode and ships with cleaner navigation, yet the off-switch has overshadowed the cosmetic upgrades. For a slice of users frustrated by aggressive AI integration across consumer software, the move positions Firefox as the only major browser explicitly built around opt-out by default rather than opt-out by accident.

The push toward AI-free browsing is not isolated. In April, Brave launched Brave Origin, a paid build of its browser sold for a one-time $60 fee — free on Linux — that strips out Leo, the Brave Wallet, Rewards, the integrated VPN, Tor windows, telemetry, and every other extra. The release uses Privacy Pass blind token technology to ensure the purchase cannot be linked to the buyer's device. Demand grew from years of viral debloat tutorials, and the company simply productized the workflow. For a cold wallet crowd allergic to data leakage, the existence of a paid no-AI browser tier signals a market segment large enough to monetize.

The backlash has not emerged in a vacuum. Chrome has been quietly distributing a 4GB Gemini Nano model to user machines that resists removal, drawing scrutiny from privacy researchers and power users alike. The browser also recently scrubbed its public disclosure promising that data processed by Gemini Nano would remain on-device and never reach Google's servers. That deletion, paired with the silent install footprint, has hardened opposition to baked-in machine learning across mainstream consumer apps. Competing projects including Dia, Opera Neon, and Comet have raced in the opposite direction, building agentic browsers that read tabs and chat with users — a bet that not everyone shares.

On the other side of the Atlantic, the AI debate took a stranger turn. Argentina's Ministry of Human Capital launched the Gemelo Digital Social initiative on May 22, presented by President Javier Milei as a paradigm change in social policy. The system is pitched as a virtual replica of Argentine society that ingests state and private data to simulate the impact of social programs before they reach citizens. The stated objective is moving the government from a reactive state to a predictive state capable of modelling poverty trajectories, subsidy outcomes, and human capital development from childhood through adulthood — the first national-scale application of the digital-twin concept to social policy.

The unveiling collapsed within hours. The official promotional video circulated by the presidency contained grammatical errors, an AI-generated avatar of Minister Sandra Pettovello speaking in her place, a Singaporean flag inserted in place of Argentina's, and a visible Amazon AWS logo left embedded in the footage. Social media mockery was immediate and intense, with critics noting that a system designed to predict the future of an entire society could not survive its own launch announcement. The aesthetic failures distracted from substantive policy questions, framing the rollout as another example of generative tooling deployed without basic quality controls inside a state institution.

Opposition lawmakers filed formal information requests demanding details on data sourcing, model architecture, and oversight, while independent privacy specialists warned that the project lacks any published governance framework. The core concern is straightforward: aggregating government and private datasets into a single predictive model — without legal guardrails — risks becoming the substrate for algorithmic surveillance at national scale. There is currently no public audit mechanism, no documented limit on data retention, and no clear answer to how synthetic citizens will or will not feed downstream policy decisions. Parallels to blockchain debates around DAO governance and on-chain transparency are not lost on observers.

The thematic arc running through both stories is the same: 2026's dominant fight is no longer whether AI will be everywhere but who decides where it goes and on whose terms. Browser vendors are discovering that opt-out is a feature worth charging for, while governments are discovering that flashy AI announcements without governance scaffolding invite immediate political cost. The privacy-first instincts that built much of the DeFi ecosystem are spreading into mainstream consumer software, and the demand for transparent, user-controlled tooling — the same impulse driving Bitcoin self-custody — is reshaping product roadmaps from Mozilla's offices to Buenos Aires.