CERT-In Issues Safety Guidelines for Using Generative AI Tools Like ChatGPT, Gemini, Grok:

The Indian Computer Emergency Response Team (CERT-In) has released an advisory highlighting the increasing security threats associated with generative AI tools such as ChatGPT, Gemini, and Grok. With the rapid adoption of these tools in daily operations, CERT-In emphasizes the need for caution and responsible usage to prevent misuse or data compromise.

Key Threats Involving Generative AI Models

CERT-In outlines multiple vulnerabilities that could affect users and organizations relying on AI tools:

Data Poisoning: Inserting malicious data into training sets to influence model behavior.

Adversarial Attacks: Slightly altered inputs can trick models into giving incorrect or biased outputs.

Model Inversion: Extraction of sensitive training data via model queries.

Model Stealing: Replication of proprietary models through continuous probing.

Prompt Injection: Malicious prompts bypass AI filters to produce harmful outputs.

Hallucination Exploitation: Using false AI-generated content to deceive or scam users.

Backdoor Attacks: Triggering harmful behaviors using hidden commands embedded in models.

Best Practices for Safe AI Usage

Best Practice	Description
Use Trusted AI Applications	Only install AI tools from verified sources to avoid malware or spyware risks.
Protect Sensitive Information	Never input personal or confidential data into AI chat tools.
Manage Access Rights	Regularly review AI app permissions, especially in business environments.
Verify AI-Generated Content	Cross-check facts from trusted sources before use.
Restrict AI to Support Roles	Avoid using AI for critical decisions in medical, legal, or business areas.
Enable Security Features	Use strong passwords and two-factor authentication for AI accounts.
Maintain Anonymity Where Needed	Use anonymous accounts and avoid sharing identifiable details.
Check for Plagiarism	Ensure content generated by AI doesn’t violate copyright laws.
Monitor for Misuse	Watch for signs of phishing, deepfakes, or scam attempts using AI platforms.

CERT-In’s Emphasis on Responsible AI Adoption

The advisory serves as a timely reminder that while generative AI tools offer powerful capabilities, they also introduce new risks. Responsible usage, proper access control, and vigilance against misuse are essential to maximize benefits while minimizing threats.