A new AI created by the company Anthropic, is sparking global cybersecurity concerns at a mytho-logical level.The AI company claimed they are doing a favour for the world by withholding the software, Mythos, that would allow anyone, anywhere (with an internet connection, of course) — to hack into any application.This new AI is trained using a new generation of graphics processing units (GPU), a type of chip capable of processing vast amounts of data simultaneously.It is capable of a level of coding only surpassed by "the most skilled humans.".Adding salt to the wound of human capability, it finds and exploits vulnerabilities in software, in some cases within code faults humans haven't been able to crack for years.Anthropic recalls when the AI solved a bug in OpenBSD, an operating security system, that was 27 years old.Although software engineers were unable to solve the glitch for nearly three decades, Mythos solved it with no sweat. In another case, a 16-year-old of an unsolvable bug in FFmpeg, a software program used to encode and decode video, seemed like nothing to Mythos, which solved it. .Anthropic explains Mythos detects vulnerabilities — potential weaknesses, or ways for hackers to potentially exploit a system.How much more efficient is the AI tool at coding in comparison to humans, you may be wondering?While it takes hours for Mythos to code something, it would take weeks for humans to develop similar code.Mythos has found thousands of "high severity" software vulnerabilities, according to Anthropic..The company recently launched its new "Project Glasswing," which gives free access to $100 million worth of usage of the software, where they can run as many security scans as they want with the AI. This is just the beginning, as Anthropic predicts AI coding capabilities will likely advance quite quickly over the next few months.Although the software is not planning on being publicly available at the level that open source AI models are — like Grok, ChatGPT, or Claude, it says it will eventually want to scale users' ability to use the software for cybersecurity purposes. Now, you may be wondering, "AI like this has got to pose international security risks" — and indeed, this does seem to be the case..So, what are they exactly?Reported by the New York Times, for US rivals such as China and Russia, the US development of Mythos could mean security consequences for falling behind.The only country outside of the US who has access to the AI is Britain, and even in the UK the Bank of England governor warned Anthropic may have "crack[ed] the whole cyber-risk world open.” The UK's AI Security Institute published a report on their experience with the application, it said coding tasks that previous AI models could not complete a year ago — Mythos could complete 73% of the time..However it stated it could not be sure Mythos could attack "well defended systems" but said it was "at least capable of autonomously attacking small, weakly defended and vulnerable enterprise systems where access to a network has been gained. "The US government has even gotten involved in discussions surrounding the US technology — at the end of April, Dario Amodei, Anthropic’s chief executive, met with members of the Trump administration after some had noted the potential for the AI to wreak havoc on computer systems. Reuters reported last week the Pentagon would be deploying the AI cybersecurity model to detect code vulnerabilities within the US government.The European Union (EU) has met with Anthropic at least three times to discuss gaining access, but the company has yet to provide it since both have not come to an agreement..According to a report last week by The Parliament, Anthropic has restricted the EU's access to the technology.Anthropic as of now has provided access to Mythos to over 40 companies, including Apple, Cisco, Google, JPMorganChase, and Microsoft.As of February, Anthropic was worth an estimated $380 billion, however according to the Financial Times , the company could soon be worth $900 billion as it has just agreed to a $30 billion investment to be finalized at the end of May.
