LLM-based AI appears to be taking on a growing weight in technology stacks that support corporate security. LLM developers are stepping up efforts to support the use of their models for cybersecurity, and more security companies are deploying LLMs across their main product lines.
OpenAI unveiled a plan to gradually expand access to GPT-5.4-Cyber, a version with enhanced cybersecurity capabilities, to users who have completed identity verification. OpenAI shifted its cyber risk response strategy from limiting what the model can do to verifying who can access its most sensitive functions. The move contrasts with Anthropic, which has allowed access to Mythos Preview only to about 40 organisations. Anthropic says the unreleased Mythos model is so strong at finding and exploiting security vulnerabilities that broad disclosure would be risky.
OpenAI accelerates expansion of cyber security-focused model GPT-5.4-Cyber.
As hackers can use AI to quickly find gaps in IT systems and launch attacks, security companies are also speeding up efforts to deploy AI to find weaknesses and build defences. TrendAI, the enterprise cybersecurity business unit of global security company Trend Micro, formed a strategic partnership with Anthropic. Through the collaboration, TrendAI will embed the Claude model across its platform to strengthen agentic workflows, automation and AI-native security operations, and pursue threat research to identify vulnerabilities in AI systems and infrastructure. Cisco is in talks to acquire Israeli AI agent security startup Astrix Security, according to reports.
AI spear versus AI shield: who will win the cybersecurity war?
TrendAI partners with Anthropic, embedding the Claude model in its security product lineup.
Cisco pursues acquisition of AI agent security startup Astrix.
South Korean security company AMCLab stressed that in an environment where attackers can use AI, a zero-trust security strategy that focuses on preventing spread on the assumption that hackers have already penetrated systems will be key to sustainable security. It said microsegmentation in particular would be a keyword.
[D2People] “In the AI era, zero-trust security is more important. Blocking lateral movement prevents damage.”
The government is also moving to prepare for a spread of hacking as AI becomes more advanced. The Ministry of Science and ICT held an emergency issue-sharing meeting with related industries over the launch of projects by Anthropic and OpenAI to use their latest AI models for cybersecurity. The National AI Strategy Committee also discussed security threats related to “Claude Mythos” and plans to improve rules for on-premises security software in the financial sector.
Government prepares measures to respond to ‘Mythos’ shock, holds emergency meeting with security firms.
Government holds issue-check meeting with telecom carriers and platforms over Anthropic ‘Mythos’ threat.
AI strategy committee activates security task force for ‘Claude Mythos response’.
Industry moves on post-quantum cryptography, which supports pre-emptive responses to scenarios in which quantum computing could neutralise existing encryption systems, are also taking more concrete shape.
Global company Cisco plans to unveil PQC technology soon and later disclose quantum networking technology. Digital certificate management company Sectigo launched Private PQC, a feature that lets companies issue and manage PQC SSL/TLS certificates in their existing certificate operating environments.
“In the era of probability computing driven by AI, a big bang is coming for internet infrastructure.”
Sectigo launches Private PQC to support testing PQC certificates in existing workflows.
The article also summarises moves by companies in South Korea and abroad around AI.
Private Technology signed a memorandum of understanding with IT infrastructure and information security consulting company T&D Soft to cooperate on information security business. The two companies plan to pursue wide-ranging cooperation spanning consulting and solutions to expand joint business opportunities in the public sector as N2SF and zero-trust adoption accelerates.
Private Technology and T&D Soft cooperate on N2SF and zero-trust business.
Jong-heon Yang (양종헌), head of the offensive division at S2W, introduced implications drawn from analysis of AI-specific attack methods and real cases at a recent seminar, including input manipulation, integrity compromise, agent interaction and synthetic identity. “If existing security was about preventing system bugs, the core of AI security lies in defending the gaps in ‘cognitive logic’,” he said.
S2W executive says security in AI era must shift from blocking to continuous vulnerability management.
AI-based cybersecurity startup Artemis raised $70 million in funding. Artemis’ product continuously monitors activity across an enterprise, including logins, cloud activity and apps, and learns normal patterns. When it detects anomalies, it provides a clear situational explanation of what is happening instead of fragmentary alerts and automatically executes actions such as locking hacked accounts. Application security posture management company Apiiro launched Apiiro CLI, a command-line interface that directly embeds security into AI-based software development workflows.
Artemis raises $70 million to automatically detect and respond to AI-based cyber attacks.
Apiiro unveils CLI that embeds security features into AI coding agents.
Cloudflare launched Cloudflare Mesh, a private networking service that integrates multi-cloud environments into a single secure network. Cloudflare Mesh supports securely connecting internal systems such as servers, databases and developer tool environments to the outside without opening company firewall ports.
Cloudflare unveils Mesh, a security network for AI agents, aiming to go beyond VPN limits.
Click Here For The Original Source
