AI giant Anthropic on Wednesday introduced the world’s most powerful model ‘Cloud Mythos’, but it was not launched for the common user. The company says that this model itself works responsibly, but its power can be extremely dangerous in the wrong hands. Banks, hospitals, electricity grids and government systems can all become its targets. Anthropic has currently given it to 12 big companies like Amazon, Google, Microsoft and more than 40 organizations under Project Glasswing, so that they can first improve the security of their systems. The name Project Glasswing is derived from a butterfly with transparent wings. It is not visible even when it is in the open. Just like software flaws remain hidden. Anthropic raised Rs 850 crore. Rs 34 crore to credit and open source institutions. Has promised to help. After 90 days, the company will tell how many flaws were detected in the model. Historical – The flaw was not caught in 50 lakh tests, it found it for the first time. In the pilot, the model discovered thousands of serious cyber flaws in many operating systems and web browsers. A 27-year-old flaw was discovered in OpenBSD, the most secure operating system. 16 year old flaw in video software FFmpeg that was not found even after 50 lakh scans. It caught on the first time. This model also showed the way to gain control of the entire system by adding many flaws in the Linux kernel. Benchmarks – 100% score in cyber testing, 93.9% score in coding This model achieved 100% success in the toughest cyber security tests. 93.9% in coding benchmarks while the previous model’s score was 80.8%. It scored 83.1% in the CyberGym test, versus 66.6% of the previous model. Let us tell you that with the launch of this model, the annual income of Anthropic has also increased more than three times to about Rs 2.5 lakh crore. CEO Amodei said – There is a lot of danger from Mythos model. We created Mythos not for cyber security but for coding. But the side effect of the power of coding was that it became equally adept at cyber attacks. If it’s in the right hands, this technology could create a more secure internet and world than ever before. But if it goes wrong then there is a lot of danger.
Source link
[ad_3]
Claude Mythos Model; 100% score on cyber scale: Anthropic creates powerful AI but halts launch; Reason: Fear of cyber disaster if it falls into wrong hands
