The Ministry of Electronics and IT has urged central ministries to delay the use of AI models from OpenAI and Anthropic due to cybersecurity concerns.
New Delhi, India Jul 13, 2026 ALN: A department under the Ministry of Electronics and Information Technology (MeitY) in India has reportedly instructed central ministries to pause the deployment of artificial intelligence (AI) models developed by prominent US-based companies, OpenAI and Anthropic. This directive comes amid rising concerns over cybersecurity risks associated with advanced AI systems, particularly in sensitive government functions.
The decision to halt the deployment of these AI models follows proposals made by representatives from OpenAI and Anthropic, who approached several ministries to discuss the potential use of their AI technologies for cybersecurity and related functions. According to reports, the memorandum issued by MeitY does not impose a permanent ban on the use of these models; instead, it raises concerns regarding the timing of their deployment and cautions against the premature implementation of such advanced systems.
While the specific number of ministries contacted by the AI companies remains unclear, it has been reported that among those approached, the finance ministry sought clarification on the potential use of agentic AI and OpenAI’s models. In a detailed six-page letter, the finance ministry proposed an examination of the deployment of GPT-5.5 for cybersecurity purposes, outlining various potential applications of AI in enhancing cybersecurity measures. However, the department ultimately rejected this proposal in a memorandum issued last week.
This situation unfolds against a backdrop of increasing scrutiny regarding the cybersecurity risks posed by sophisticated AI technologies. Union IT Minister Ashwini Vaishnaw has recently advocated for the establishment of a new AI law, emphasizing that the existing IT Act is outdated and does not adequately address the rapid advancements in AI technology. In a related statement, Finance Minister Nirmala Sitharaman has urged entities in the financial sector to maintain heightened vigilance against the potential threats posed by AI, noting that such technologies could automate vulnerability discovery and facilitate malicious interference with source code, thereby making cyberattacks more rapid and difficult to detect.
"The tools of attack are evolving at high speed, and the tools of defence must evolve even faster," Sitharaman remarked, highlighting the urgent need for adaptive cybersecurity measures in the face of advancing technologies.
Advanced AI models, such as those developed by OpenAI and Anthropic, possess the capability to scan code, identify software flaws, and assist in responding to cyber incidents. However, these capabilities present a dual-use dilemma: the same vulnerabilities that AI can identify for defenders could also be exploited by malicious actors. Both OpenAI and Anthropic have acknowledged this risk and have implemented various safeguards for their models that possess advanced cybersecurity capabilities.
A case in point is Anthropic’s Claude Mythos Preview, which is specifically designed for cybersecurity applications. This model is only available to select organizations through the initiative known as ‘Project Glasswing.’ In June, Anthropic expanded access to this program to approximately 150 organizations across more than 15 countries, including India. Participants in this project can utilize the Mythos Preview model to identify critical zero-day vulnerabilities and rectify them before they can be exploited by attackers. However, access to this model was initially restricted after testing revealed its potential to rapidly uncover thousands of vulnerabilities, raising concerns about possible misuse.
The initial cohort of organizations granted access to ‘Project Glasswing’ included major tech firms such as Amazon, Google, Microsoft, Apple, NVIDIA, as well as cybersecurity companies like CrowdStrike and Palo Alto Networks, along with the UK’s AI Security Institute. This illustrates the competitive landscape in which advanced AI technologies are being developed and utilized, raising questions about equitable access and the implications of technological dependency.
The recent developments regarding the deployment of AI models have reignited discussions about technological sovereignty in India. In June, Anthropic temporarily suspended access to Claude Mythos 5 and Claude Fable 5 for foreign nationals in order to comply with US export controls. Although these restrictions were lifted shortly thereafter, the incident highlighted the complexities and challenges surrounding access to advanced AI models, particularly when such decisions are influenced by regulations and policies outside of India.
For India, this poses significant challenges as the country aims to leverage AI technologies in strategic areas while advocating for greater technological sovereignty. In pursuit of this goal, the Indian government has committed a substantial investment of ₹10,372 Crores to the IndiaAI Mission, which encompasses initiatives such as subsidized computing infrastructure, startup financing, and support for the development of homegrown foundation models. The government has also expressed support for domestic AI developers, including companies like Sarvam AI and the BharatGen consortium.
Despite these efforts, the framework governing the deployment of frontier AI models across government remains ambiguous. Existing regulations concerning data localization, cloud infrastructure, and the management of government data were not designed with the complexities of advanced AI systems in mind. This lack of clarity leaves unresolved questions regarding data processing locations, retention of information by providers, and the hosting of underlying AI models. As India navigates the evolving landscape of AI technologies, establishing a robust and clear regulatory framework will be crucial for ensuring the secure and effective deployment of AI in government functions.
In conclusion, the pause on the deployment of AI models from OpenAI and Anthropic reflects a cautious approach by the Indian government in the face of emerging cybersecurity threats. As the landscape of AI continues to evolve, it will be imperative for policymakers to balance the potential benefits of these technologies with the associated risks, all while striving for greater technological independence and security in critical sectors.
To learn more about the latest developments in Startup Policies & Regulations, stay updated with our exclusive reports and analyses on AiLensNews.
