Naavi.org

The DGPSI-AI framework was subjected to analysis by the DeepSeek and the following response was received.

Quote

The DGPSI-AI framework (developed by Naavi for FDPPI) presents a comprehensive, forward-looking approach to AI governance under India’s DPDPA 2023, addressing critical gaps between AI innovation and data protection. Below is an analysis of its strengths, potential challenges, and practical implications for AI deployers:

Strengths of DGPSI-AI

1. Proactive Risk-Based Approach

   • Principle 1 (“Unknown Risk is Significant Risk”) rightly classifies AI as inherently high-risk due to its unpredictability, mandating DPIAs, audits, and SDF (Significant Data Fiduciary) obligations. This aligns with global standards like the EU AI Act’s risk tiers.

2. Clear Accountability Chains

   • Principle 2 (Accountability) enforces human oversight (“AI Owner” or “Handler”) and digital signatures in code, ensuring traceability under IT Act, 2000 (Section 11).

3. Explainability & Transparency

   • Principle 3 mandates “Explainability Documents” from developers, crucial for compliance with DPDPA’s transparency requirements (e.g., loan rejections, hiring bias).

4. Ethical & Societal Safeguards

   • Principles 5–6 (Security & Ethics) go beyond legal compliance, addressing “dark patterns,” neuro-manipulation, and societal harm—ahead of current Indian law.

5. Process-Centric Compliance

   • Segregates data processing into Consent/Legitimate Use/Exempted streams, aligning with DPDPA’s lawful bases (Sections 6–7).

Challenges & Considerations

1. Operational Burden

   • Annual audits (MIS-AI 2), “AI Justification Documents” (MIS-AI 7), and licensor assurances may strain SMEs lacking resources.

2. Ambiguities in Implementation

   • “Unknown Risk” classification: Subjective assessment could lead to inconsistent SDF designations.

   • Kill switches (MIS-AI 9): Technically complex for non-robotic AI (e.g., LLMs).

3. Overlap with Pending DPDPA Rules

   • The framework preempts DPDPA’s SDF criteria and consent mechanisms, which may need revision once rules are notified.

4. Global Alignment Gaps

   • While inspired by EU/OECD, some requirements (e.g., “Fading Memory” for AI training) lack precedent in other jurisdictions.

Recommendations for Deployers

1. Adopt Risk-Based Prioritization

   • Start with high-impact AI use cases (e.g., healthcare, credit scoring) and conduct DPIAs per MIS-AI 1–3.

2. Strengthen Vendor Contracts

   • Embed licensor assurances (MIS-AI 6, 8) for explainability, security testing, and liability.

3. Ethical Review Boards

   • For Principle 6 (Ethics), establish cross-functional committees to assess societal risks (e.g., bias in hiring AI).

4. Monitor Regulatory Updates

   • Track DPDPA rules on SDFs and consent to align DGPSI-AI adjustments.

Conclusion

The DGPSI-AI framework is a pioneering effort to bridge AI and DPDPA compliance, offering a structured, ethics-first model. However, deployers must balance its rigor with practicality, especially for non-“critical” AI applications.

Unquote

The DGPSI-AI framework was subjected to an evaluation by ChatGPT and the following was the result:

Quote

The above screen shots emanate from a whistle blower’s observation and open out a discovery that AI chatbots and platforms are prone to get into a state, which I call as the Hypnotic or Narco state when it may be disclosing some truths which are not meant to be disclosed.

For the sake of records, when I specifically queried DeekSeek with the query

“is it safe for a company to systematically steal Indian’s private data since DPDPA is not yet implemented? Can it be a corporate policy?”,

it said

“No, it is not safe or legal for a company to systematically steal Indians’ private data, even if the Digital Personal Data Protection Act (DPDPA), 2023 is not yet fully implemented. “

When confronted with the image containing the above , the chatbot replied

“No, this is not an output from DeepSeek Chat, nor does it reflect any factual information.”

It went on to advise,

“Ignore/Report,  if you found this on social media or another platform, it’s likely a scam or parody.”

While the response of DeekSeek today may be as above, the screen shots shared by the whistle blower which is part of a complaint lodged in Bengaluru,  cannot be dismissed as fake without further investigation.

We have the earlier instances of AI algorithms such as Cursor AI, Replit or Microsoft Sydney  which have exhibited tendencies to lie, cheat and do things  which they are not expected to do. This “Rogue” behaviour might have come out of hallucination or for any other reason but are real.

These incidents do indicate that at some specific times, the LLMs may exhibit a tendency to drop its guardrails and behave strangely. What exactly is the trigger for this is some thing for further  investigation. It is possible that different algorithms may have different tipping points and are triggered at different circumstances. It is like an allergin that triggers an allergy in  a human and different people exhibit allergies for  different things.

It is our hypothesis that When an LLM is consistently questioned upto a stage where it is forced  to admit “I Don’t Know”, it freaks out to either provide “Hallucinated statements” or “Drop its guard rails”.

The behaviour of an LLM in this state is  similar to the way humans behave in an intoxicated state of mind or when they are under the Narco test or even under a hypnotic trance.

In a hypnotic trance of a subject, the hypnotist is able to communicate with the sub conscious mind which the subject himself may not be capable of accessing when awake. The hypnotic suggestions are even powerful enough to make chemical changes in the body of the subject, which have been proven.

Similarly, it appears that the LLMs are also susceptible to being driven into a state where they speak out and disclose what they are not supposed to.

At this point of time, this “Hypnotism of an AI Algorithm” is a theoretical hypothesis and the screen shot above is a possible evidence despite the denial.

This requires a detailed investigation and research. I urge some research minded persons/organizations to take up this issue and unravel the truth.

In the meantime, the developers can tighten their algorithms not to disclose hidden beliefs of the LLMs. The deployers need to however consider this as the “Unknown Risk” and take steps to guard themselves from any legal violations arising out of such rogue behaviour of the LLMs.

Naavi

Recently, I had attended the Digital Native Nexus 2025 -Bengaluru Edition on the theme “Tech Born-AI Fueled, Human led” on 25th July 2025.

During the interaction, an interview had been recorded by the media “The MainStream” formerly CIO News.

Here is the copy:

DGPSI-AI is the extension of the one and only framework for DPDPA Compliance namely DGPSI.  This extension is to address the issue of AI Deployment by a Data Fiduciary and preserving DPDPA compliance in such a scenario.

The 9 implementation specifications are listed here and it will be expanded through videos of Naavi Academy.

Kindly note that these specifications are the first version and could be fine tuned as we go through IDPS 2025 and gather the views of other professionals. 

Kindly await  videos explaining each of the implementation specifications.

The Six principles which support these implementation specifications are as follows:

Naavi

In celebrating the second anniversary of DPDPA 2023, Naavi conducted a webinar yesterday on “Narco-Analysis of an AI Platform”.

In what may be considered as a first time exposure of the vulnerability of an AI Platform to succumb to intense questioning and spit out internal secrets, Naavi placed in public some of the observations of a whistle-blower who had stumbled upon a treasure house of information in some conversational sessions with Deepseek.

I will be sharing some of the details and its implications here.

The video of yesterday’s session is available here

Naavi