RSAC 2026 (cybersecurity conference)RSAC 2026RSAC 2026conference / summitdàhuì大会focused on / centered onjùjiāo聚焦AI (artificial intelligence)AIAIAI agent / autonomous agentzhìnéng tǐ智能体securityānquán安全threatwēixié威胁，expertszhuānjiā专家warnedjǐnggào警告defensefángyù防御system / frameworktǐxì体系urgently needsjí xū亟需to restructure / to rebuildchónggòu重构。
March 26, 2026 at 10:00 PM
New Words
News in English
Science & TechnologyRSAC 2026 Highlights AI Agent Security Threats, Experts Warn Defenses Need Rebuilding
The RSAC 2026 cybersecurity conference in San Francisco focused on the new security threats posed by autonomous AI agents. Experts warned that traditional defenses can no longer address machine-speed attacks and called for a fundamental overhaul of enterprise security architecture.
The 35th RSAC cybersecurity conference was held in San Francisco from March 23-26. This edition's theme focused on data security chaos in the artificial intelligence era.
Participating experts pointed out that the rapid proliferation of autonomous AI agents is dismantling traditional human-centric identity verification and access control frameworks.
So-called shadow agents—autonomous programs running without enterprise security team approval—are regarded as one of the primary sources of current data breach risks.
Security researchers warned that attackers can exploit compromised AI platforms to implant malicious agents, completing reconnaissance and vulnerability exploitation at machine speed.
At the conference, the industry reached three consensus points on defense strategies: protecting agents from external intrusion, preventing agents from harming external systems, and achieving machine-speed detection and response capabilities.
Multiple security vendors demonstrated AI-versus-AI defense solutions, positioning real-time machine learning detection as a scalable line of defense.
Deepfake attacks and AI supply chain poisoning were also classified as major practical threats currently in operation.
Analysts believe that given the rapid evolution of AI capabilities, enterprise security architecture needs fundamental restructuring, rather than merely layering patch-style protections on existing frameworks.
More in Science & Technology
ChinaZhongguó中国AIréngong zhìnéng人工智能modelsmóxíng模型captureszhànlǐng占领marketshìchǎng市场
Chinese AI Models Gain Market Share as GLM 5.2 Rivals OpenAI and Anthropic at Low Cost
UBTechYōubìxuǎn优必选U1U1U1humanoid robotrénxíng jīqìrén人形机器人launchesfābuō发布pricejiàgé价格
UBTech Launches U1 Ultra-Bionic Humanoid Robot Series Starting at 17600 Dollars
Apple (company)Píngguǒ苹果to file a lawsuit / to sueqǐsù起诉OpenAI (AI company)OpenAIOpenAI，to accuse / to chargezhǐkòng指控it / the company (formal)qí其to steal / to misappropriateqièqǔ窃取hardwareyìngjiàn硬件trade secrets / confidential business informationshāngyè jīmì商业机密。
Apple Sues OpenAI Alleging Hardware Trade Secret Theft
Tencent (tech company)Téngxùn腾讯inzài在WeChat (messaging app)Wēixìn微信to launch / roll outtuīchū推出artificial intelligenceréngōng zhìnéng人工智能assistantzhùshǒu助手Xiaowei (AI assistant name)Xiǎowēi小微。
Tencent Begins Rolling Out AI Agent Xiaowei to WeChat Users