AI Policy

Guruswami Advisory Guruswami Pty Ltd | ABN 11 695 354 020 Last updated: 11 March 2026

Guruswami Advisory uses artificial intelligence in our research, analysis, and content development. As an AI advisory firm, we hold ourselves to the same standards we recommend to clients.

1. How We Use AI

We use AI systems in the following ways:

  • Research and analysis. We use AI to explore research questions, synthesise literature, and identify patterns across large bodies of evidence. All findings are independently verified before inclusion in client deliverables or published content.
  • Content drafting. AI assists in drafting articles, briefings, and advisory documents. All published content is authored, reviewed, and approved by Paul Nevin. AI accelerates the drafting process. It does not replace editorial judgment.
  • Adversarial testing. We use AI to challenge our own analysis, stress-test recommendations, and identify weaknesses in our reasoning before presenting to clients. This practice is central to our methodology.
  • Infrastructure research. We operate our own AI inference infrastructure (the Chakra Cluster) for hands-on evaluation of models, architectures, and deployment patterns. This informs our technical advisory work.

We do not use AI to make advisory decisions. Every recommendation reflects the professional judgment of our advisory team.

2. Client Data and AI

No client data is processed by third-party AI services.

  • Client names, documents, financials, strategies, and engagement details are never submitted to any external AI platform, including commercial APIs.
  • When AI-assisted analysis is required for client work, it is performed on infrastructure we control, using models we have evaluated for data handling, logging, and retention.
  • We treat all client information as confidential by default, consistent with our Privacy Policy and engagement agreements.

This commitment applies regardless of whether a client requests it. We do not rely on vendor assurances about training data exclusion. We verify where data goes.

3. Content Disclosure

Published content on guruswami.com, including Insights articles and white papers, is produced with AI assistance and human oversight:

  • Claims and statistics are independently verified against primary sources. Every cited figure links to its source.
  • Analysis and recommendations reflect the professional judgment of our advisory team, informed by AI-assisted research but not generated by it.
  • Infographics and visual assets are produced with AI-assisted generation tools, then reviewed for accuracy and consistency.

We do not disclose AI involvement on a per-sentence basis. The above describes our standard workflow for all published content.

4. Our Own Governance

We apply the same governance principles we recommend to clients:

  • Bias awareness. We actively seek disconfirming evidence for our own positions. Our published framework, the Enterprise Bias Pipeline, was developed from our own research practice.
  • Model evaluation. We do not rely on a single AI system. We evaluate multiple models and cross-reference outputs where the stakes warrant it.
  • Transparency over performance. When AI-assisted analysis produces an uncertain result, we disclose the uncertainty rather than present a confident narrative.
  • No autonomous decision-making. AI informs our work. It does not approve deliverables, set scope, or communicate with clients.

5. Regulatory Alignment

This policy is informed by:

  • DTA AI in Government Policy (mandatory from June 2026) — principles of transparency, accountability, and human oversight.
  • ASIC REP 798 — findings on AI governance gaps in regulated entities, including the need for disclosure and fairness policies.
  • APRA CPS 230 — operational resilience requirements that extend to AI-dependent processes.
  • Australian Privacy Principles — data handling obligations under the Privacy Act 1988 (Cth).

We review this policy against evolving regulatory requirements and update it as standards change.

6. What We Decline

We will not use AI in ways that compromise the quality or integrity of our advisory work:

  • We will not present AI-generated analysis as independent research.
  • We will not use AI outputs to support a predetermined conclusion without adversarial review.
  • We will not process client data through AI systems that lack adequate data handling controls.

Changes to This Policy

We may update this policy as our practices and the regulatory landscape evolve. The "Last updated" date at the top reflects the most recent revision.

Contact

For questions about how we use AI in our work:

Email: [email protected] Location: Canberra, ACT, Australia