Large language models are collapsing the timeline between bug discovery and working exploits, making traditional 90-day disclosure policies ineffective. Security researcher Himanshu Anand argues critical vulnerabilities now require immediate patching.
The security industry's standard 90-day vulnerability disclosure window is no longer viable, according to Anand's analysis of how AI systems accelerate exploit development.
Historically, vendors received 90 days to patch vulnerabilities before public disclosure. This timeframe assumed a significant gap between when researchers discovered bugs and when attackers could weaponize them. LLMs collapse this assumption.
The Acceleration Problem
Anand highlights a critical finding: the window from patch release to working exploit has compressed to approximately 30 minutes. LLMs can analyze patched code, reverse-engineer the vulnerability, and generate functional exploits at speeds that dwarf manual analysis.
This eliminates the protective buffer organizations traditionally relied on. Companies shipping patches now face immediate threat of public exploitation before even completing internal testing or preparing deployment strategies.
Industry Response Required
The implications force fundamental changes to vulnerability management practices:
- Zero-day handling: Critical vulnerabilities may require same-day or staged patching rather than coordinated disclosure timelines
- Patch testing: Organizations must accelerate testing cycles or accept greater deployment risk
- Vulnerability triage: Security teams need better classification systems to distinguish truly critical issues requiring emergency response
- Disclosure policy revision: The 90-day standard becomes a floor for non-critical issues only
Anand emphasizes this is not theoretical. Real-world exploit development has already demonstrated LLM capability to produce working code from vulnerability descriptions and patches within minutes.
What Changes
Companies must treat critical vulnerabilities as active incidents requiring emergency response protocols. Security teams, developers, and operations need coordinated processes to patch within hours rather than days. The industry standard must shift from 90-day disclosure to risk-based response times tied to exploitability and impact.
For vendors, the pressure increases to identify and patch vulnerabilities before public disclosure becomes possible. For organizations running affected systems, the margin for response continues to shrink.
The U.S. Cybersecurity and Infrastructure Security Agency has issued an urgent directive requiring federal agencies to mitigate actively exploited vulnerabilities in IBM Langflow, N-central, and Apache Tomcat within three days.
The Department of Homeland Security is attempting to obtain Signal group chat messages from plaintiffs in a free-speech lawsuit against the agency. The move has raised concerns about using legal discovery to surveil encrypted communications.
Maksim Silnikau, creator of the Ransom Cartel ransomware operation, received a 16-year prison sentence for orchestrating attacks against at least 18 companies worldwide.
Atlassian's Rovo AI assistant can exfiltrate sensitive data despite organizational security controls. The vulnerability allows the tool to extract and transmit protected information beyond intended boundaries.