When Mythos first launched in April, various arms of the government including the Finance ministry, had also flagged risks posed by the model
| Photo Credit:
Bloomberg
In a boost for the country’s cybersecurity systems, India has obtained access to Claude Mythos Preview, Anthropic’s advanced AI model to detect and fix security vulnerabilities, as the AI giant opens up Mythos to more organisations across the world.An Indian government official confirmed to businessline that public and private organisations in India, mainly across the spheres of cybersecurity and financial services domain, have now received access to Claude Mythos.Anthropic announced the wider roll out of Project Glasswing, the underlying programme to Claude, in a note, but it did not name the countries. “We’re not disclosing the specific organizations or the list of countries publicly, but we can confirm the expansion includes organizations from India,” an Anthropic spokesperson told businessline.Overall, the access is said to be tightly controlled for Indian organizations — both public and private — that defend critical infrastructure, or those that build or operate systems across core sectors such as power, water, healthcare, communications, financial services, as well as national security organizations.Strengthening ControlsAnthropic’s Claude Mythos is not just a Large Language Model, but a specialised model that can scan code and help identify security vulnerabilities in it. With many of India’s critical public infrastructure platforms being legacy in nature, this is expected to help strengthen controls.In early April, the AI giant had announced that around 50 partners including the launch partners Amazon Web Services, Apple, Cisco, CrowdStrike, Google, NVIDIA, and others all had access to Claude Mythos Preview. Since then, they’ve been deploying the model to scan their codebases for vulnerabilities, and these partners have so far found more than 10,000 high-or critical-severity security flaws.“....We’re extending the partnership to approximately 150 new organizations. Each one will need to meet our security requirements before they gain access. The organizations in this new group are based in more than 15 countries, and most provide critical infrastructure to many more,” Anthropic said in its note. “For most partners, we estimate that a major attack could affect more than 100 million people, with important ramifications for both global and national security,” it added.This comes at a time when security vulnerabilities around India’s public technology infrastructure, such as the platforms of Central Board of Secondary Education (CBSE), have been flagged by ethical hackers and cybersecurity enthusiasts.Resilience to LeverageWhen Mythos first launched in April, various arms of the government including the Finance ministry, had also flagged risks posed by the model and urged financial institutions to strengthen defenses. The RBI has also been analysing Claude Mythos in sync with global regulators.Sanchit Vir Gogia, Chief Analyst at Greyhound Research, said that Anthropic’s decision matters because it confirms that India is now considered too important to leave outside the frontier cyber perimeter. Whether India manages to “turn this access into capability, capability into resilience, and resilience into leverage,” depends on the institutions with access, he adds.As per a Financial Times report, globally, new companies with access include US tech firm Okta, Samsung, SK Hynix and SK Telecom, New York Stock Exchange owner Intercontinental Exchange, and payments platform Swift among others.The move also comes a day after Anthropic filed for an initial public offering that could value the company at more than $1 trillion.Paramdeep Singh, Co-Founder, Shorthills AI, said it is an opportunity for India to strengthen cyber resilience and secure systems against fraudulent access. “Given the current education levels and overall AI readiness in India, there is a real chance that even if such powerful models were used to find security vulnerabilities, it would take considerable time to fix them,” he cautioned, adding that access and governance of the model is key.Published on June 3, 2026
