Colombo | 23 August 2026:
The Centre for Policy Alternatives (CPA) notes recent news reports to the effect that Sri Lanka is considering the use of Artificial Intelligence (AI) in legislative drafting, beginning with a pilot project linked to the proposed Virtual Asset Service Providers (VASP) Bill. Government officials are reportedly exploring the potential use of AI tools for “drafting, validating, and producing” (the basis for the distinction between these three terms is unclear) legislative texts in Sinhala, Tamil, and English, while also discussing data readiness and safeguards. While technological innovation in public administration and the legislative process should not be dismissed, the proposal raises important questions that extend beyond the capabilities of AI itself and go to the heart of governance in Sri Lanka.
The attraction of AI in legislative drafting is obvious. The institutional actors of the legislative process, within the executive and even more so in Parliament, are often under-resourced and face increasing demands for speed, complexity, and multilingual accessibility. AI tools have the potential to assist with legal research, identify inconsistencies, compare draft provisions with existing statutes, generate summaries, and automate a range of repetitive and time-consuming tasks. In principle, these capabilities could contribute to a more efficient and responsive legislative process.
However, the central question is not whether AI can improve drafting efficiency but whether the institutional environment within which it is proposed to operate is capable of ensuring its responsible, transparent, and accountable use.
Sri Lanka’s governance record across multiple administrations led by all parties has been characterised by persistent concerns regarding policy inconsistency, weak institutional capacity, inadequate consultation, limited transparency, and poor accountability. CPA has since its inception criticised the legislative process for the hurried introduction of bills, insufficient public scrutiny, and limited stakeholder engagement. The NPP Government’s partisan, strategic, unprincipled, and populist behaviour in relation to the ongoing Twenty Second Amendment to the Constitution Bill saga vividly demonstrates that it is no different from the older parties of government when it comes to sharp practice and mendacity. Yet again, the system has changed those who were elected to change the system. Against this backdrop, claims that AI will improve legislative quality should be approached with caution.
There is a risk that AI becomes another technological solution deployed to compensate for deeper governance deficiencies rather than addressing them. Poor policy choices do not become better policies simply because they are drafted more quickly or what may seem like passable English. Likewise, legislation that has not undergone adequate public consultation does not become more democratic because it was generated with the assistance of sophisticated software.
Indeed, many of the risks identified internationally are likely to be amplified in governance environments where institutional safeguards are weak and political culture does nothing to disincentive zero-sum behaviour on the part of the government in power (or for the time being popular). AI systems are known to generate errors, produce inaccurate outputs, and reproduce biases embedded in their training data. Experts consistently emphasise that human review and oversight are essential safeguards against these risks. Yet in Sri Lanka, one must ask whether the same public institutions that frequently struggle with basic implementation, coordination, and regulatory enforcement are adequately prepared to supervise highly technical AI systems in a meaningful way.
Transparency presents another significant concern. Legislative drafting is ultimately an exercise of public power. Citizens are entitled to know not only what laws are proposed but how they are produced. If AI-assisted tools are used in legislative development, important questions arise regarding the data on which such systems are trained, the assumptions embedded within them, the criteria used to evaluate their outputs, and the extent to which public officials rely on machine-generated recommendations. Without robust transparency requirements, AI risks introducing an additional layer of opacity into an already opaque policymaking process.
Questions of accountability are equally important. When flawed legislation produces unintended consequences, responsibility should be traceable to identifiable decision-makers. However, governments around the world are increasingly confronted with the problem of “automation washing,” where decisions that are ultimately political are presented as objective outputs of technological systems. Such practices can weaken accountability by obscuring who is responsible for mistakes, poor judgements, or harmful outcomes. In the legislative sphere, there is a danger that AI-generated recommendations could acquire an unwarranted aura of neutrality or expertise, despite reflecting underlying assumptions, data limitations, and design choices made by humans. The introduction of AI into the administrative, policy-making, and legislative process could therefore exacerbate the problem of overcentralisation embodied in the Sri Lankan presidential system, the catastrophic lack of accountability of which led to national bankruptcy and social hardship from which we are still recovering. The NPP was elected in 2024 to abolish this system but as with every party coming into government promising abolition of the executive presidency, it appears that that will once again be a broken promise.
There are also significant concerns regarding data governance and security. Legislative drafting frequently involves sensitive policy deliberations, confidential communications, and internal government documents. While proponents argue that AI can enhance aspects of security, these benefits depend on robust institutional safeguards, continuous monitoring, and strong data governance frameworks. Sri Lanka’s track record does not inspire confidence that such safeguards can simply be assumed to exist.
For these reasons, AI should not be viewed as a substitute for competent public administration, rigorous legal expertise, or democratic accountability. At best, it is a tool that may support legislative drafters in carrying out specific technical functions. At worst, it may provide governments with a convenient means of accelerating lawmaking while reducing transparency, diffusing responsibility, and evading accountability. That the recent initiative, while hosted by the Legal Draftsman’s Department, appears to be driven by the panjandrums in charge of the “digital economy” should also be cause for the serious concern that (executive) technocracy rather than (parliamentary) democracy will be the source of legitimacy relied upon by the government in selling this policy to the public. That concern is heightened in the emerging order of the Sri Lankan political economy as a cosy combine between private sector plutocracy and leftwing party-government.
The introduction of AI into legislative drafting should therefore proceed only on the basis of clearly articulated safeguards. These should include mandatory human review, public disclosure of AI use, independent audits for bias and accuracy, clear accountability mechanisms, strong data protection standards, and meaningful opportunities for public scrutiny. Most importantly, technological innovation should not be allowed to distract from the more fundamental challenge facing Sri Lanka: the need to strengthen the quality of democratic governance itself; and in this context, the increasing ephemerality of the NPP’s promise of a new constitution to abolish the executive presidential system.
The real problem confronting lawmaking in Sri Lanka has never been a shortage of drafting tools. It has been the absence of consistent transparency, accountability, institutional competence, and public participation. AI cannot solve these problems. Indeed, if introduced without adequate safeguards, it may simply reproduce them in a more technologically sophisticated form. As Pope Leo XIV warned in Magnifica Humanitas, there is every danger with the unwise use of AI of building a Babel in the place of a New Jerusalem. That remains a very real prospect in Sri Lanka under NPP governance.