Manifest V3 was supposed to fix the extension ecosystem. Google announced it, delayed it, extended deadlines, and finally enforced it - and the result is a quieter mess than the one it replaced.
The original problem was real. Browser extensions, particularly in Chrome, had accumulated extraordinary permissions. Many could read every page you visited, intercept network requests, and exfiltrate data with no meaningful checkpoint between installation and execution. The ad-fraud and credential-stealing operations that ran through compromised or deliberately malicious extensions were not theoretical. They were documented, repeatable, and profitable.
Manifest V3 tackled some of this. Extensions now face tighter restrictions on how they handle network requests - the webRequestBlocking API that ad blockers and security tools relied on was replaced with declarativeNetRequest, which shifts control over what gets filtered from the extension to the browser itself. Google framed this as a security improvement. Ad blocker developers framed it as deliberate sabotage of their tools. Both things can be true.
The Permission Model Is Still Broken
What V3 didn’t fix is the part that causes the most damage in practice: the acquisition problem. Extensions are bought and sold. A developer builds something genuinely useful, grows an install base of a few hundred thousand users, then sells to a buyer who has no interest in the original product. The new owner pushes an update. Permissions silently expand. Data starts flowing somewhere new.
Chrome does prompt users when an extension requests new permissions after an update, which is something. But the bar for what counts as a meaningful permission expansion is low, and most users click through without reading. The Web Store review process catches some malicious extensions before publication, but post-acquisition updates have historically slipped through - sometimes for months.
Firefox’s approach isn’t dramatically better, but Mozilla has been more willing to revoke extensions retroactively and communicate why. Chrome’s sheer market share makes its permissiveness a systemic issue rather than an edge case.
The Real Leverage Nobody Uses
Operating systems already sandbox applications by default on mobile. iOS and Android have made it structurally difficult for an app to silently access your contacts, location, or clipboard without surfacing a prompt at the moment of access. Desktop browsers have never matched this. Extensions still get granted sweeping access at install time, often across all sites, indefinitely.
A per-site, per-session permission model - where an extension has to re-earn access to sensitive contexts rather than holding it forever - would be a genuine structural fix. It would also break a lot of extensions that users find genuinely useful, which is probably why it hasn’t happened.
The extension ecosystem is too useful to abandon and too compromised to trust. V3 moved the furniture around. The house still has the same foundation.