Google enabled “Site Isolation”, which runs a different browser process for each internet domain as a defense against Meltdown and Spectre attacks, in Chrome 67
Google has secretly enabled a security feature called Site Isolation for 99% of its desktop users on Windows, Mac, Linux, and Chrome OS.
Context & Ripple Effects
Site Isolation is Google's structural answer to Meltdown and Spectre: rather than patching each side-channel as it appears, Chrome 67 redraws the browser's trust boundary so every internet domain runs in its own process, making it far harder for code on one site to read another site's data. The rollout to roughly 99% of Windows, Mac, Linux, and Chrome OS desktop users was done quietly — no prompt, no opt-in — which says Google considered the architectural change worth shipping silently at scale.
This is the foundation stone for a decade of Chrome hardening covered here before: Google later extended Site Isolation to Android login pages in Chrome 77, and by 2023 was building on the same origin-boundary logic with an opt-in IP Protection feature that masks addresses through proxies. Isolation came first because every later defense assumes sites can't reach each other.
First-order effects
- For the ~99% of desktop users covered, Spectre-style attacks that abuse shared processes to steal passwords or session data across sites lose their primary browser vector — the attacker's script now sits in a process that cannot map another domain's memory.
- Google absorbs the cost silently: more processes mean higher memory use and rendering overhead on every tab-heavy machine, traded away without asking users, which is why the enablement reads as 'secret'.
Second-order effects
- Rival browser makers face pressure to match per-site process isolation or explain why their engines leave cross-site data exposed to the same class of attacks — isolation stops being a differentiator and becomes table stakes.
- Web developers building sites that embed heavy third-party content now operate inside a multi-process reality, where cross-origin iframes and widgets carry real process-spawn costs they must design around.
Third-order effects
- If the pattern holds, the browser consolidates as a compartmentalized operating system in miniature — each origin a sandboxed tenant — which is precisely the substrate later Google proposals like IP Protection presuppose when they route 'qualifying traffic' through proxies on top of hardened boundaries.
The trend: Browsers are being rebuilt around the origin as a hardware-enforced trust boundary, with Google shipping isolation first and layering network-level privacy defenses on top of it.