Anthropic is tightening the clock on security disclosure. Beginning January 1, 2027, the company will shorten its coordinated disclosure deadlines to 30 days for high- and critical-severity vulnerabilities, and 45 days for other issues β the window a vendor gets to fix a flaw before Anthropic can publish it.
Coordinated disclosure runs on a bargain: researchers report flaws privately, vendors fix them, and the public hears about it afterwards. The traditional timelines assumed human-speed discovery. AI systems now find vulnerabilities far faster, and the same tools that find a flaw defensively can find it offensively.
Shorter deadlines pressure vendors to patch at machine speed β and quietly admit that the old calendar no longer matches how security actually works.
The background
Coordinated disclosure is the security world's long-standing truce. A researcher who finds a flaw reports it privately, the vendor gets a fair window to fix it, and only then do the details become public. The classic windows β often around 90 days β were calibrated for a human world: human researchers finding flaws one at a time, and large organisations patching on slow, careful cycles.
AI breaks both halves of that calibration. Capable models can now discover vulnerabilities far faster than human teams once did, scanning code at a scale no research group could staff. And once a flaw is public, the same class of tools can help attackers turn a description into a working exploit. The comfortable gap between 'found' and 'weaponised' is shrinking from both directions.
Why it matters
Shorter deadlines move real costs onto vendors. A 30-day clock for high- and critical-severity flaws means patch development, testing and rollout have to happen at something closer to machine speed β a serious operational demand for companies whose software runs hospitals, banks and infrastructure. Anthropic is effectively declaring that the old calendar was written for a threat environment that no longer exists.
Disclosure norms also tend to move as an industry. When one major lab shortens its timelines, others face pressure to explain why theirs are longer, and vendors find the strictest policy becoming the de facto standard they must meet. A single company's policy change can quietly become everyone's deadline.
What happens next
The new clocks start on January 1, 2027, which gives vendors a year to rebuild their patch pipelines for the shorter windows. The things to watch are whether vendors actually meet the deadlines, whether other AI labs adopt similar timelines, and what happens the first time a serious flaw goes public on day 30 with no fix ready β the moment the new bargain gets its first real test.
Common questions
What is Anthropic changing about vulnerability disclosure?
From January 1, 2027, Anthropic will shorten its coordinated disclosure deadlines. Vendors will have 30 days to fix high- and critical-severity vulnerabilities its AI finds, and 45 days for other issues, before the findings can be published.
Why are the deadlines getting shorter?
Because AI systems now discover vulnerabilities much faster than human researchers did. The old timelines assumed human-speed discovery; with machines finding flaws β and able to help exploit them β the safe window for patching has shrunk.




