Mythos Will Create a Vulnerability Surge and Collapse the Exploitation Window.

No - and this is the most important misunderstanding to correct. The US Commerce Department’s export control directive on June 12, 2026 removed defensive access for organizations outside the US, including India’s regulated enterprises that had been onboarded through Glasswing’s 15-country expansion. The offensive capability was unaffected. The model leaked on the day of announcement — Anthropic could not keep it contained among 60 handpicked partners for even 24 hours. The UK AI Security Institute has independently confirmed that OpenAI’s GPT-5.5 matches Mythos on expert-level cybersecurity benchmarks and is broadly available today without export restrictions. Anthropic’s own public statement cited GPT-5.5 parity as evidence that the jailbreak triggering the directive was not uniquely dangerous. The restriction removed a defensive tool from Indian organizations while doing nothing about the diffused capability. Anthropic itself estimates that comparable capability will be available as open-source models within 12 to 18 months.

Previous automated scanners match software versions against known CVE databases. Mythos-class AI does something categorically different: it reasons about code the way an expert human researcher would, identifying novel attack paths, chaining multiple weaknesses into working exploits, and generating functional proof-of-concept code - all without documentation, without authorization, and without fatigue. It found a 27-year-old flaw in OpenBSD that human researchers missed across decades of review. It identified a vulnerability in wolfSSL that would have allowed forging TLS certificates across billions of IoT and industrial devices. It spotted an FFmpeg flaw that automated tools missed across 5 million prior scans. The difference is the depth of reasoning, not the speed of pattern-matching. This is why CERT-In titled its blueprint specifically “Blueprint for Reducing Exposure and Defending against AI-Assisted Vulnerabilities Exploitation” — because the threat class is qualitatively different from what existing controls were designed to address.

CERT-In’s directions and the blueprint apply to all service providers, intermediaries, data centers, corporates, and government entities operating in India. There is no sector exemption. That said, the risk concentration and regulatory consequence are highest for organizations simultaneously subject to multiple frameworks: BFSI organizations face CERT-In plus RBI, SEBI, IRDAI, and DPDP Act simultaneously. IT services companies and GCCs face CERT-In as an India-specific obligation on top of their global parent’s compliance requirements. Telecom operators carry Critical Information Infrastructure designation. Healthcare organizations face Significant Data Fiduciary candidacy under DPDP. Every regulated Indian organization is in scope. The question is not whether the mandate applies but how wide the compliance gap currently is.

The 12-hour remediation mandate for Known Exploited Vulnerabilities on internet-facing and crown-jewel systems is the hardest requirement to operationalize, and it is hard for structural reasons that cannot be fixed by effort alone. The average Indian organization’s change management process — raising a change request, convening a change advisory board, testing in staging, scheduling a maintenance window, and deploying — takes two to four weeks under normal conditions. For vendor-locked systems running Finacle, FlexCube, or legacy COBOL mainframes, the patch timeline extends to months, because the fix belongs to the vendor. CERT-In’s 12-hour window is not achievable through human-speed workflows under any realistic staffing model. Meeting it requires either autonomous patch deployment with AI-scored production safety, or patchless compensating controls — WAF rules, service disablement, network isolation — that satisfy the containment requirement without requiring a maintenance window. Both capabilities are available in the Qualys platform. Neither is available in a scan-and-ticket workflow.

Incident reporting is one requirement among dozens in the blueprint. Compliance requires meeting all of the following simultaneously: 12-hour KEV remediation for internet-facing systems; 6-hour incident reporting with structured evidence to CERT-In, RBI (separately), and SEBI (if applicable); continuous exploit validation with demonstrable proof that attack paths are closed; AI system governance across all 16 Section 12 requirements including shadow AI discovery, prompt injection testing, and AIBOM generation; xBOM supply-chain transparency (SBOM, AIBOM, QBOM, CBOM) per CERT-In’s Technical Guidelines v2.0; 180-day log retention within Indian jurisdiction; and regular assessment by CERT-In empanelled auditors. Meeting the reporting requirement while leaving the remediation, validation, AI governance, and xBOM requirements unaddressed is not compliance. It is partial compliance with the easiest requirement.

xBOM is the family of “bill of materials” artifacts CERT-In now mandates — Software Bill of Materials (SBOM), AI Bill of Materials (AIBOM), Quantum Bill of Materials (QBOM), and Cryptographic Bill of Materials (CBOM). Each one is a structured inventory of a specific category of component in your technology stack. CERT-In requires them because supply chain attacks — where an attacker compromises a dependency your software trusts rather than your software directly — are India’s second-largest breach vector at 17% of all incidents. Without an xBOM, you cannot know which third-party components you depend on, cannot assess whether a compromise in one of those components affects you, and cannot satisfy CERT-In’s auditors that your supply chain is governed. Qualys CSAM generates all five xBOM artifact types continuously from the same asset intelligence layer that powers vulnerability management — so the inventory is always current, not a point-in-time snapshot that is obsolete by the time an auditor reviews it.

Yes, and this is one of the most underappreciated aspects of the blueprint. Section 12 mandates AI asset inventory as its first requirement — specifically because most organizations do not know how much AI they have deployed. Shadow AI is the dominant pattern: employees deploy ChatGPT integrations, GitHub Copilot, AI-powered customer service tools, and automation workflows without IT authorization or security review. IBM’s 2025 India breach data found shadow AI was a top-three breach cost driver adding ₹17.9 million per incident on average, and fewer than 42% of Indian organizations have any policy to detect it. If your employees are using any AI tools — which they almost certainly are — Section 12 applies to you. CERT-In’s mandate requires you to discover what you have, govern it, test it for prompt injection, and generate AIBOM before you can demonstrate compliance.