News · 2026-11-17
Anthropic and Google set shorter clocks for publishing security flaws
Anthropic will shorten its vulnerability disclosure windows for new reports in 2027, while Google Project Zero plans a staged reduction to a 30-day deadline in 2028. Both organizations argue that faster discovery changes the security calculus, but their policies set publication clocks rather than prove fixes will reach users sooner.
Key facts
- Anthropic’s new policy begins January 1, 2027: 30 days for high and critical issues and 45 days for other issues.
- Google Project Zero uses “Full 90” now, plans Full 60 on February 2, 2027, and Full 30 on January 3, 2028.
- Both retain separate seven-day treatment for actively exploited flaws.
- The primary sources are Anthropic’s update and Project Zero’s policy.
Coordinated vulnerability disclosure starts when a researcher privately informs the software maintainer about a weakness. The maintainer can develop and distribute a fix before the researcher publishes technical details. The deadline balances two risks: publishing too early can expose users before a fix is ready, while waiting indefinitely can leave users unaware and give attackers time to discover the same flaw.
Anthropic says its new window applies to reports made on or after the start of 2027; reports already in progress retain their prior deadlines. It aims to involve an external coordinator after 15 days without a maintainer response. It also reduces its post-patch wait before sharing full technical details from 45 to 30 days. These are distinct clocks. A 15-day escalation target is not a 15-day demand that a patch be ready.
Google’s timetable works differently. Under its “Full X” approach, Project Zero plans to publish technical details at the report-date deadline whether or not the issue is fixed. It removes the former extra 30-day period after a patch was released. Google uses one overall deadline across severity levels, while Anthropic differentiates its deadlines by severity. The Project Zero explanation, written by Natalie Silvanovich, describes a staged transition and the concerns vendors raised.
Anthropic argues that models can help find and repair flaws and that separate actors may rediscover a vulnerability more quickly. Google says attackers can reverse-engineer a patch, potentially making a delay after release less protective than it appears. Imagine a door manufacturer silently ships a new lock: someone comparing the old and new locks may infer how to open doors whose owners have not replaced theirs. The interval between availability of a fix and its installation is the patch gap.
The strongest counterargument comes from the maintainers Google consulted. A shorter clock can disclose unfixed vulnerabilities before organizations have improved their ability to respond; it may also push lower-priority issues ahead of more dangerous ones or make deadlines appear impossible. Anthropic acknowledges that some flaws are difficult to repair or deploy and says it will work with vendors on remediation timelines. Google’s policy allows rare publicly recorded exceptions to avoid widespread user harm.
The two organizations also treat existing reports differently from newly reported ones. Anthropic says reports already underway when its new schedule begins retain their prior deadlines. Google says a vulnerability follows the policy in effect when it was reported. That matters to a maintainer already coordinating a fix: an announced future policy does not automatically reset every pending clock. Both also distinguish issues under active exploitation, for which a shorter seven-day policy remains relevant. These exceptions are not evidence that every urgent flaw can be fixed in seven days; they describe how quickly the researchers plan to disclose information under specified conditions.
There is no measured outcome yet for these new policies. A deadline may encourage a developer to release a fix, but it does not itself update millions of devices or prove that the time to installation fell. Anthropic’s disclosure dashboard describes earlier work and human review of model-generated candidates; it predates the new policy and cannot demonstrate its effect. The practical question to watch is not merely when reports become public, but whether maintainers can validate patches and downstream users can actually install them in time.
Key questions
When does Anthropic’s shorter disclosure policy begin?
When will Google Project Zero reach a 30-day policy?
Do shorter deadlines mean every user is patched sooner?
Cite this
APA
Ground Truth. (2026, November 17). Anthropic and Google set shorter clocks for publishing security flaws. Ground Truth. https://groundtruth.day/news/google-anthropic-shorten-vulnerability-disclosure-clocks.html
BibTeX
@misc{groundtruth:google-anthropic-shorten-vulnerability-disclosure-clocks,
title = {Anthropic and Google set shorter clocks for publishing security flaws},
author = {{Ground Truth}},
year = {2026},
month = {nov},
url = {https://groundtruth.day/news/google-anthropic-shorten-vulnerability-disclosure-clocks.html}
}
Comments are replies to this story on Bluesky — reply with any Bluesky account to join in.