US President Donald Trump believes artificial intelligence should have a “kill switch” possibly to keep a check on the rapid advancements in AI and ensure those advancements do not create a scenario where humanity comes under any existential threat. In an interview broadcast on Wednesday, Trump suggested that while AI has the potential to enhance sectors like banking by making it more secure and efficient, it also brings considerable risks.
During his discussion with Fox Business Network, Trump was asked about the possibility of AI affecting the banking system’s stability. He acknowledged the risks, stating, “Yeah, probably. But it could also be the kind of technology that allows greatness in the banking system, makes it better and safer and more secure.”
Trump further indicated that there should be government safeguards on AI technology, including a potential “kill switch.” His comments were made during an interview on “Mornings with Maria,” which was recorded on Tuesday. No further details about how and if – and when – such a switch could come into action have been shared at this point in time.
The US President’s remarks followed warnings from cybersecurity experts regarding Anthropic’s new AI model, Claude Mythos. Experts have cautioned that Mythos could enhance complex cyberattacks, posing significant challenges to the banking sector, especially given its reliance on legacy technology systems. Anthropic has declined to comment on these warnings but has stated that Claude Mythos Preview will not be widely available.
Mythos is seemingly so powerful, Anthropic has tied up with some 40 plus companies who can use this new AI model, with the primary use case being identifying flaws in software that might make them prone to any cyberattacks. Anthropic, at least from the point of view of messaging and access, is practising due diligence, but it is also the same AI company currently classified as a “supply chain risk” by the US Government, which bars federal agencies from using its tools and services. Anthropic, which is led by Dario Amodei, a former co-founder of OpenAI, is contesting the directive in court.
Though it is not available to the public yet, data shared by Anthropic itself suggests that Mythos packs immense power, even in “preview” state. As an example, it was able to find a 27-year-old bug in OpenBSD while simultaneously detecting thousands of zero-day vulnerabilities in open and closed-source software. In fact, word on the Internet is that Mythos is apparently capable of hacking into any software.
The need to one-up competition is also pushing rivals like Sam Altman’s OpenAI to launch its own version of Mythos. OpenAI is calling it GPT 5.4-Cyber which is essentially a souped-up version of GPT 5.4 designed specifically for catching software flaws. Like Mythos, GPT 5.4-Cyber is also available to select users, though OpenAI is yet to announce any major corporate tie-ups, while Anthropic has the likes of Apple and Amazon who are onboard using Mythos.
