Anthropic claims that Mythos is skilled at cybersecurity tasks, specifically combing large codebases to find vulnerabilities that even human reviewers have not discovered for decades. File.
| Photo Credit: Reuters
Anthropic PBC, the firm that develops the AI tool Claude, announced on Wednesday (July 1, 2026) that it would resume access to Fable and Mythos, two of its most powerful Large Language Models (LLMs), later in the evening. The models had been pulled globally after the United States government put out an export order restricting their use for anyone who was not a US person. The June 12 ban came just days after Indian organisations (it is yet to be announced which) obtained access to Mythos, the more advanced model which has not been made available to the public. Anthropic claims that Mythos is skilled at cybersecurity tasks, specifically combing large codebases to find vulnerabilities that even human reviewers have not discovered for decades. Fable is also highly capable on such tasks, but Anthropic has erected protections against many use cases. For instance, one researcher using Fable told The Hindu that biology-related tasks were restricted on the model, something Anthropic said is important to prevent the creation of bioweapons.“Working closely with the government, we trained an improved safety classifier that targets and blocks the behavior described in the report” by Amazon that triggered the US response, Anthropic said in a statement. “Users will be notified if a request to Fable 5 is blocked, and the request will instead be sent to Opus 4.8,” the strongest model that was available for much of last month.Sovereignty pangsWhile the Union government has been acutely aware of the concentrated U.S.–China dynamic in frontier AI technology development, the abrupt cutoff to Mythos led to heightened discontent. Reports indicated during the ban that the government would leverage its support to Sarvam AI, the Bengaluru-based $1.5 billion sovereign AI startup with two open source models under its belt, to get a small equity stake in the company.On Tuesday (June 30, 2026), a senior official reiterated what the U.S. government indicated at the Pax Silica summit last week: if access has been given, it will not be taken away. The official stated that India’s focus on AI — unlike western economies which are dealing with employment-related concerns — would be focused on gaining maximum utility out of AI, and delivering “impact” widely. Published - July 01, 2026 11:08 pm IST
