Cybersecurity News Roundup – 13 October

A compilation of cybersecurity news trending over the last 48 hours:

Cybersecurity

• The Shai-Hulud worm has reached AI infrastructure through a Tensorlake compromise: credential-stealing malware was caught minutes after an npm release, but the impact is still unknown. theregister
• CrowdStrike found what may be a bank hacker’s CV in exposed AI logs. The suspect, believed to be a Chinese speaker, used Claude Code and an agentic pentesting tool called ARTEX against South Korean lenders. Separately, South Korea’s president has ordered a probe after a series of financial data breaches. theregistertechrepublic
• Ukrainian drone strikes on Yandex datacenters took cloud and payment services offline, and network connectivity fell to 30% of normal. theregister
• The US disrupted Chinese hacking tools while seven governments warned that PRC spies are stealing sensitive data worldwide. The NCSC-led advisory names Integrity Technology as a facilitator for the Flax Typhoon state actor. theregistertechtarget
• Critical infrastructure: Booz Allen tests showed AI breaching industrial controls in 16 minutes. The Register reports AI agents can already move robotic arms and falsify OT operator screens. techwireasiatheregister
• Patch now: Citrix NetScaler has a new flaw rated 9.5, with exploitation status unknown. The FBI confirms FortiBleed attacks are ongoing, with more ransomware groups joining in. Attackers hijacked top-level domains and issued fake certificates for Google and other organisations. Technology news and analysis | The Register +2
• TechTarget looks at CISOs fighting escalating deepfake fraud, two years after the Arup attack. techtarget

Agentic AI and its risks

• AWS AgentCore’s security was undone by a prompt that asked for credentials. Contributing factors were tokens in metadata, weak VM isolation and broad permissions. theregister
• Meta released its Muse Wonder models for personal and professional agents. TechTarget says it faces a trust gap with enterprises. siliconangletechtarget
• Google brought agentic AI to Gemini, starting with business customers. TechCrunch
• Anthropic says it will act to defend critical infrastructure and open-source projects from AI, and expects attackers to hold the advantage for the next two years. theregister

BFSI and regulation

• Singapore’s central bank wants every fintech AI use case independently reviewed, and won’t accept third-party failures as an excuse. Singapore is also weighing tougher safeguards for high-risk agentic AI. theregistertechwireasia
• Evident’s AI index ranks JPMorgan Chase and Capital One highest for AI maturity among banks. HSBC could cut 70% of its human wealth advisers as AI takes over client service. ciodivetechtarget
• The UK ICO says model developers need a lawful basis for processing personal data, and that claiming to benefit humanity doesn’t qualify. techtarget

India angle

• The US has barred Microsoft, Adobe and major IT firms from the green card program for skilled foreign workers. India’s tech giants have so far shrugged off the visa changes. TechCrunchtheregister
• Elon Musk has stepped up his attacks on Ambani over the delayed Starlink India launch. TechCrunch
• Airtel Business launched IQ Gateway, a DPDP-compliant platform that combines fraud protection with network insights. techtarget

Infrastructure, deals and market

• Microsoft debuted its multichiplet Maia 300 inference accelerator. AMD unveiled Helion racks for AI agents, with partner systems due in November. siliconangletheregister
• Cloudflare acquired Deno to improve its Workers programming model. TechCrunch
• Global PC shipments fell 20% as AI-driven component shortages pushed prices up. Samsung is eyeing an $80B quarterly profit on memory pricing. theregistertheregister

Sources: TechCrunch Enterprise, The Register, SiliconANGLE, CIO Dive, TechRepublic Security, TechTarget News, TechWire Asia, GeekWire, CIO.com