India’s cybersecurity agency, CERT-In, has issued fresh security guidelines asking organizations to fix critical vulnerabilities that face the internet within 12 hours, where it’s possible and practical. This call is being made as worries keep mounting that threat actors are now leaning more on artificial intelligence tools and large language models (LLMs) to speed up cyber incidents and, in a sense, automate parts of the exploitation process.
CERT-In says that AI-assisted cyber activity is massively cutting the time attackers need to spot weaknesses, map what services are publicly exposed, abuse weak credentials, and get into systems that were left unprotected. The agency also cautioned that today’s cyber threats are turning out to be not only quicker but also more scalable and increasingly automated.
Meanwhile, researchers point out that cybercriminals are already using AI for things like phishing wording, malware creation, exploit scrutiny, attack-surface mapping, and even vulnerability discovery. The upshot is that attackers can squeeze their planning phase, and they may slip past several older, traditional defensive measures.
CERT-In also pointed out the risks that come from AI systems themselves, too. In other words, threat actors can mess with AI-enabled platforms via prompt injection, model manipulation, training data poisoning, data leakage, jailbreaking methods, and even by compromising the orchestration pipeline. If this kind of thing goes wrong, it can seriously weaken the confidentiality, integrity, and reliability of the AI systems that organizations rely on day to day.
In its set of recommendations, CERT-In said organizations should work with an “assume breach” mindset and also tighten their operational readiness. The agency stressed that rapid detection, containment, and recovery abilities matter a lot, especially because AI-driven attacks are getting more and more sophisticated.
The blueprint also suggests moving toward Zero Trust architecture, using defense-in-depth in a layered way, cutting down exposure to internet-facing services that are vulnerable, and locking down cloud environments, APIs, identities, and the software supply chains. Organizations are told to keep continuous monitoring, verify security controls, and stay aware with full visibility across AI-enabled systems and third-party dependencies.
One of the more notable recommendations is about strict patching timelines, and you know, not just “eventually.” CERT-In said that known exploited vulnerabilities that hit internet-facing systems should, where possible, be patched within 12 hours. For critical external issues, the remediation should happen within a day, and for high-severity weaknesses, they should be handled within five days, but this is always with risk prioritization in mind, so it is not exactly one-size-fits-all.
If security patches are not ready right away, organizations are basically told to lean on temporary mitigations like access restrictions, workload isolation, stronger monitoring, API protection, or even turning off vulnerable features until the real fixes land.
The agency also really emphasized proactive cybersecurity work such as penetration testing, red teaming, vulnerability assessments, independent audits, and software bill of materials validation (SBOM) to lower software supply chain exposure, and yeah, to keep the dependency story cleaner.
The new guidance kinda reflects this global worry that AI technologies lower the bar for cybercriminals, and they can launch attacks at this kind of unprecedented speed and scale, almost too fast. Security experts say organizations can no longer just depend on traditional reactive defense models, and they need to pivot toward continuous monitoring, quick patch management, and operational resilience.
Cybersecurity companies like IntelligenceX keep stressing proactive threat intelligence, vulnerability management, AI security monitoring, and rapid incident response, too, as AI-assisted cyber threats get more advanced and harder to spot.
Top comments (0)