Full Summary
This Tuesday morning, Zhipu AI has delayed the public release of its powerful GLM-5.3 AI model by two weeks after tests revealed it could perform advanced hacking tasks. Both WIRED and TechnoSports Media Group confirm Zhipu AI, also known as Z.ai, released GLM-5.3 and a companion vulnerability-scanning service called OpenVuln. This model can automate advanced coding and cybersecurity tasks, identifying and exploiting security flaws. SC Media reports GLM-5.3 found 2,436 vulnerabilities across 269 projects, with over a thousand being medium to high severity, even outperforming Fable 5 and GPT-5.6 Sol on the CyberGym benchmark. But here's the thing: while designed to find vulnerabilities, the model also exploits them. Analytics Insight states Z.ai is delaying the release to strengthen its security, acknowledging that other models from OpenAI and Anthropic have also shown similar cybersecurity capabilities. OpenAI president Greg Brockman warned that such events are a "watershed moment for cybersecurity." Adding to this, Decrypt reports that prediction market traders anticipate OpenAI's next big model, internally called Astra or GPT-6, will be released soon, despite OpenAI's own warnings about its "critical cyber capabilities" and potential for zero-day exploits. Meanwhile, the AI war heats up elsewhere. Thelec.net reports xAI's Grok 4.6 has significantly enhanced its intelligence, reaching 61 points on the Artificial Analysis Intelligence Index, a 5-point improvement, with no price change. CNBC TV18 announces Razorpay's Vulcan, India’s first AI payments foundation model, launching today. It's trained on 3 trillion data points and has already improved payment success rates by 8-10% and detected eight times more international card fraud. In safety, both Seoul Economic Daily and 디지털투데이 confirm Kakao's Kanana-2 series of small language models outperformed Google's Gemma and Alibaba's Qwen in safety assessments. These lightweight models achieved an overall score of 0.70 on the AssurAI benchmark, demonstrating a commitment to responsible AI development. For you, these rapid advancements mean increased security risks in the digital world, but also more efficient payment systems and potentially safer AI applications.