HYDERABAD: With AI being the buzzword in every field, its usage is increasingly gaining currency and acceptance. The state’s Commercial Tax Department (CTD) officials have been unofficially using AI in their work for drafting discrepancy reports, notices, orders, and other proceedings. They have also been using it for analysing, translating and summarising case records.
The Circular No 1/2026 issued by the department on August 18 in its bid to balance the confidentiality of taxpayer information and obligations under the Digital Personal Data Protection Act, 2023 welcomes responsible adoption of AI tools to enhance speed, consistency and quality. However, while focusing on what is barred, it fails to give actionable guidelines to officials.
It further errs by shifting the legal, disciplinary and criminal burden onto individual field officers without establishing an institutional infrastructure, budget allocations and risks delaying adoption of technology for better efficiency.
The circular imposes absolute bans on inputting any taxpayer data into third-party AI platforms, integrating departmental databases (GSTN, Back Office, BIFA, e-way bill) via APIs or browser extensions, and using personal devices or accounts for processing official records. The circular only permits generic legal research in the abstract and language or grammatical polishing of drafts.
The circular while permitting officers to frame queries in generic (wholly hypothetical) terms, opines that the taxpayer information can be leaked easily. But it doesn’t suggest the way to navigate safely without the risk of data leak while warning, their usage amounts in law to disclosure of confidential taxpayer information to a person not authorised to receive it.
Ambiguous circular
The circular is ambiguous on the upper and lower thresholds of AI usage. It, however, warns officials of personal disciplinary action under the TS Civil Services (CCA) Rules, 1991, criminal prosecution under Section 133 of the TGST Act and Section 5 of the Official Secrets Act, 1923, and civil liabilities under data protection laws. But as the boundary between a “generic abstract query” and an “inadvertent disclosure of taxpayer context” is subjective, officers face risk of being punished.
Commenting on this, an official states, “Why will any official risk using AI tools, he or she will try to avert personal exposure? This will affect productivity,” he avers.
The department, while acknowledging the existence of free or paid platforms, mentions the possibility of platforms “expressly authorised in writing... by the Commissioner,” but it does not address who should foot the bill on such usage.
It may be remembered here that free, public AI platforms use user inputs for model training, save user data on foreign servers, the situation the circular seeks to avoid. So, officials will need secure enterprise-grade models that ensure data isolation, zero-retention policies, and dedicated security. This makes usage of paid commercial subscriptions or institutional licensing a necessity.
But there is no clarity on provision for budget allocation, software reimbursement, or departmental enterprise licensing. Expecting officers to independently access compliant AI solutions without institutional procurement mechanisms makes safe, authorized adoption practically unworkable.
It further identifies the department as a “Data Fiduciary” under Digital Personal Data Protection Act, 2023. The paragraph 14(c) in the circular warns unapproved disclosures will amount to reportable personal data breaches.
Moreover, while addressing usage of cloud transmission, API integrations, data localisation, and browser vulnerabilities, what comes out starkly is the absence of CERT-In compliance standards and government cloud security architecture is a notable technical oversight, another official said on condition of anonymity.
Compliance burden
By not providing for rollout of an in-house or government-hosted AI environment, it puts the burden of the entire operational and compliance burden on the individual end-user, the officials. This is despite the fact that the GST Council, Central Board of Indirect Taxes and Customs (CBIC) and several progressive state commercial tax administrations (like ADVAIT and BIFA modules) are actively deploying AI and machine learning for predictive audit targeting, automated mismatch detection, and behavioral analysis.
The officials TNIE spoke to fear that the circular could lead to technological stagnation in the department. Telangana could fall behind other states that are equipping their workforce with secure, high-yield digital tools. The buzz in the department is that the circular issued ostensibly to regulate use of AI is itself a product of AI usage. This, they aver, can be asserted by the language used typical of AI tools.