[DigitalToday reporter Chi-gyu Hwang] AI-powered cyberattacks are surging, and concerns are spreading as hacking incidents continue to break out in countries around the world.

Recently, as AI has sharply improved its mathematical abilities, warnings are emerging that internet encryption technology could be put at risk. Some experts appear to be leaving open the possibility that AI could crack even the core encryption technologies that protect online bank accounts and databases. Quantum computers have long been seen as the biggest variable threatening encryption, but some views hold that this premise is being shaken by advances in AI.

· UK Financial Conduct Authority urges overhaul of encryption systems in light of AI threats: "We must start now"

In South Korea, consumers and government authorities appear to be on high alert as hacks hit one after another, from major commercial banks to secondary financial institutions, investment firms and cryptocurrency exchanges.

· Hacks hit Daishin Securities, Bithumb and Mirae Asset Global Investments in succession; securities industry on 'security emergency' · Financial authorities conduct emergency inspections of hacks in the digital asset sector; push to strengthen security · 4.4 million cases of access from North Korean IPs targeting the financial sector detected; authorities request security checks

The hacking of enterprise software company Douzone Bizon also appears to be having a significant impact. Regarding the Douzone Bizon case, the Ministry of Science and ICT urgently convened a review committee for incident investigations. It judged the incident to qualify as a major case and concluded that a joint public-private investigation team needs to be formed.

· Douzone Bizon says breaches occurred in backup services and SaaS; scale still being assessed

At large U.S. banks, unauthorized external AI agents operating inside customer accounts are being detected one after another. Some agents also showed signs of trying to move independently of user instructions.

· Large U.S. banks on 'emergency' over AI agents accessing customer accounts

Moves by South Korean companies toward AI security are also accelerating.

Igluu Corporation will expand its cyber security AI agent brand AiR (AI Road) into a three-product lineup: AiR mate, AiR works and AiR clue. As part of that effort, it will also launch an AI agent orchestration platform that designs, operates and manages AI agents within an organization. Jun-hee Lee (이준희), CEO of Samsung SDS, also stressed AI security as one of the future growth engines, based on partnerships with global AI companies. PFSO AI will soon unveil an AI security platform that comprehensively manages various information leakage paths when companies use AI and provides visibility across overall AI usage.

· Igluu Corporation expands AiR security AI agent lineup; also to launch AI orchestration platform · Samsung SDS CEO Jun-hee Lee: "Cooperation results with global frontier AI companies becoming visible in AI security business" · PFSO AI to soon launch AI security platform focused on AI visibility

South Korean companies that do not specialize in security are also actively entering the AI security market. Polaris Office will launch a generative AI security product based on global security company Netwrix's data loss prevention (DLP) solution, Endpoint Protector, and move to target the enterprise market. Genon, a generative AI solutions company, launched its enterprise AI security solution GenGuard.

· Polaris Office makes full-fledged push into enterprise AI security market · Genon launches AI security solution GenGuard...sets security policies in natural language

It also summarized the moves and issues surrounding security among companies in South Korea and abroad.

AI security specialist Aim Intelligence unveiled an application security assessment service linking its AI red-team solution Stinger with OpenAI's cyber security programme Daybreak. Security operations and analytics platform company Igluu Corporation obtained a patent for collecting log data without interruption and sequentially storing it even when network failures occur in cloud environments.

· Aim Intelligence unveils application security assessment service linked to OpenAI Daybreak · Igluu Corporation obtains patent for 'cloud-based log management server'...collects log information even during failures

Krafton combined Microsoft Security Copilot and Microsoft Defender XDR with its own systems and data to automate alerts, investigations and reporting.

· Krafton automates security alerts, investigations and reporting with AI...deploys Microsoft Security Copilot

Anthropic updated its security service Claude Security to continuously inspect connected code repositories without a person having to start checks each time.

· Anthropic updates Claude Security...continuous checks of code repositories

Microsoft again asked administrators of its cloud-based account and sign-in management service Entra ID to switch user authentication methods to passkeys. Fortinet expanded AI security features in its FortiGate firewall. Amazon Web Services (AWS) formally launched a security service, AWS Continuum 'code vulnerabilities', which finds and fixes code vulnerabilities. Trend AI integrated Nvidia's open-weight AI model Nemotron into its security platform Vision One Sovereign·Private Cloud (SPC).

· Microsoft to end Entra ID text and voice authentication...recommends shift to passkeys · Fortinet adds AI security to FortiGate firewall...real-time prompt inspection · AWS launches Continuum code vulnerabilities...selects only risky flaws and suggests fixes · Trend AI unveils AI security platform run in customers' data centres

Dell Technologies introduced two AI security services. The 'Dell AI Factory Security Enhancement' service supports customers in finding configuration risks by having Dell experts inspect customers' AI factories across physical, virtual and application layers. The 'AI Resilience Operations' service provides a plan to identify essential AI functions critical to operations and revive them after disruptions through application recovery policies and procedures.

· Dell expands cyber resilience lineup...also targets AI security

There was also news that the AI model connection tool LiteLLM, because of an accidentally exposed administrator key, had its deployment system taken over by hackers, resulting in the distribution of a malicious version.

· LiteLLM's deployment system hacked after accidental exposure of a single admin key

Nikesh Arora (Nikesh Arora), CEO of Palo Alto Networks, said that as model performance improves, vulnerabilities will continue to be discovered amid rising concerns over AI-driven hacking. He said it is impossible to find and fix them all at once because the discovered vulnerabilities are scattered across different areas. "If you apply many patches at once, you also don't know what side effects might occur in infrastructure," he said, offering 'hybrid patching' as an alternative.

· [Tech Insight] Air raid of AI hacking..."SOC innovation and hybrid patching are realistic alternatives"

Amazon CEO Andy Jassy (Andy Jassy) stressed the response to AI-based cyberattacks, saying, "Since AI models find far more vulnerabilities than people do, companies need to rewrite nearly all code so agents can handle it."

· "In the AI era, companies must rewrite software code so agents can patch it"

Hyun-ho Lim (임현호), head of Check Point Korea since taking the helm in November last year, made clear his intention to use an AI-first strategy to turn sentiment around in the South Korean market. The details of that turnaround can be summed up as taking the lead in the early race in the newly opening AI security market, then using that as a foothold to pursue existing companies in the firewall market. To that end, Check Point also reorganised its network firewall product lineup around AI.

· [D2People] "AI security-first results becoming visible...momentum to expand firewall share as well"

As companies race to pursue AI transformation (AX) to change their business constitution by using AI, AX is also gaining speed at the scene of cyberattacks. In the early days, hackers used AI to build tools and infrastructure needed for ransomware attacks. These days, it has evolved to the level of using AI to search through already infiltrated corporate networks and find ways to break into other accounts.

· From attack automation to big-data analysis...'hacker AX' also gathers pace