7 Comments
User's avatar
Ron Smith's avatar

Great essay, well composed and very informative. It sickens me that the US has behaved in this immoral manner, but it’s not new. I am no expert but it seems China has followed the US lead in many actions similar to this. Will we ever learn that back door behavior by governments will see light, be exposed and not only ruin government reputations, but threaten millions of lives and our collective future? With software solutions underpinning the world increasingly this concern becomes more salient everyday. I am as concerned with this short term threat to the modern world as I am with the long term threats as climate change. We seem to be longing for our own dooms.

Zhou Mo 周墨's avatar

the 48-hour rule has a famous test case: alibaba cloud found log4shell in nov 2021, told apache first, and MIIT suspended it as an info-sharing partner for six months. with AI finding bugs faster, that pressure to tell the ministry first only grows.

Mansi Kulkarni's avatar

The tension between information control and cybersecurity disclosure is particularly interesting. Effective cyber defence often depends on researchers being able to identify vulnerabilities, share information and report weaknesses without excessive institutional barriers. This raises a broader question about whether centralised control can adapt to a domain where security increasingly depends on openness and cooperation.

Roger Chua | The Long Game ♟️'s avatar

That 4% vs over a quarter patched is wild. In my experience the slow part is rarely the big firms, it's the small businesses running some old WordPress site nobody really owns anymore.

Roger Chua | The Long Game ♟️'s avatar

That 4% vs over a quarter patched is wild. In my experience the slow part is rarely the big firms, it's the small businesses running some old WordPress site nobody really owns anymore. Curious how Singapore and the rest of Asia looked in the Censys numbers.

AI Frontier Post's avatar

China patching only 4% of its vulnerable services in a week versus over a quarter in the US is stark, but Tencent capping bounties at $15,000 while Apple and Google pay millions gives researchers a clear reason to report to the state instead. A model finding a zero day for $450 means the cheap part is now discovery, not the patch.

AI Delivery Brief's avatar

The number that changes the job is minus seven days. When AI makes finding the bug cheap, the exploit window opens before the patch exists, so patching becomes the cleanup, not the plan. The control has to be pre-positioned instead: canary deploys, fast rollback, and monitoring that catches the break-in attempt, like Amazon's detection here, before the fix ships.