AI startup company Anthropic has discontinued its most advanced AI models ‘Cloud Fable 5’ and ‘Mythos 5’ worldwide. The company has taken this decision after the order of the US government, in which it was asked to ban the use of these models for foreign citizens. Now no non-American citizen including Indians will be able to use them. American officials have cited the threat of national security and cyber attacks behind this. This is the first time a government has directly banned the use of AI software rather than hardware or chips. Fear of cyber attack through jailbreak: The US government fears that through these advanced AI models, the security cordon ‘jailbreak’ i.e. digital jail can be bypassed. If this happens, they can be used to find vulnerabilities in computer systems, hack government data or carry out major cyber attacks on banking systems and critical infrastructure. According to a report in ‘The Wall Street Journal’, during a testing, Amazon researchers had used some special prompts to remove some software flaws from Anthropic’s model. This information was given to the US Commerce Department, after which strict action was taken. Company bid – The decision to ban is wrong. On this matter, Anthropic says that this decision of the government is the result of a big misunderstanding. According to the company, the flaws that were found were very limited and similar flaws can also be found by other public AI models available in the market. It is not right to ban it all over the world on such a large scale. The company said that before launching it, it had been tested for weeks with US government agencies and Britain’s AI Safety Institute, then no one had found any such major threat. The company is now in talks with the government so that this access can be restored as soon as possible. Till then users can use the rest of the company’s old AI models. There was already a dispute going on with the Trump administration. There has been tension between Anthropic and the Trump administration for the last few months. In fact, the company had refused to allow American security agencies to use its AI models in domestic surveillance and fully automatic weapons. After this, the Pentagon put Anthropic in the list of ‘supply-chain risk’. This controversy has increased at a time when the company is preparing to launch its IPO in the stock market, due to which its market value is being estimated at around 1 trillion dollars (about Rs 92 lakh crore). Zoho Founder Sridhar Vembu said – Globalization is over now. Till now America’s focus was only on preventing AI chips and semiconductor technology from reaching other countries (especially China). But this new ban on software has shocked the whole world. Expressing concern over this decision, Sridhar Vembu, founder of Indian tech company Zoho, has called it a big wake-up call for India. Vembu said that now globalization is over. India should now rapidly develop its own sovereign AI capabilities and promote open-source models instead of relying on foreign companies, because any country can block access to its technology at any time in the future. Knowledge Box: What is ‘Jailbreak’ and ‘Sovereign AI’? Jailbreak: Some security rules are set in AI models so that they do not provide information like making bombs, carrying out cyber attacks or teaching illegal activities. When a hacker or user cleverly breaks these rules and extracts restricted information from AI, it is called ‘jailbreak’. Sovereign AI: A domestic AI system developed by a country with the help of its own data, infrastructure and scientists. It is not controlled by any outside country or foreign company. ———— Also read this news… Threat of cyber attack on banking system from Mythos AI: Finance Minister Sitharaman held a high-level meeting; What is mythos and why is it dangerous? Finance Minister Nirmala Sitharaman held a high-level meeting with the heads of banks on Thursday. In this meeting, the possible threats to the banking sector from Anthropic’s ‘Cloud Mythos’ AI model were discussed. This AI model is so advanced that hackers can use it to carry out cyber attacks on the financial sector by discovering decades-old unknown flaws. Read the full news…
Source link
