Organizations are struggling to keep up with the security risks posed by AI agents, with most not treating them as identities and lacking proper governance and security measures. This oversight leaves them vulnerable to security breaches and data exfiltration. The rise of shadow AI, created by employees and business units without oversight, poses a significant risk profile.
Snowflake is addressing this issue with a framework that organizes agentic security into three layers: data, model, and agent. This approach ensures defense in depth across the full workflow. Meanwhile, Cisco is pushing further into quantum security and enterprise AI, partnering in the Quantum Communication Fieldlab Rotterdam and launching the Secure AI Factory with Equinix and NVIDIA.
AI adoption is also facing gaps despite growing investment, with many organizations struggling to move AI projects from proof-of-concept to production. Perplexity has launched Brain, a self-improving memory system that builds a context graph of an agent's work and learns overnight. However, concerns have been raised about the hype surrounding generative AI, with some warning that it could lead to a backlash against all AI research if the bubble pops.
Anthropic has taken its new Claude Fable 5 and Mythos 5 AI models offline following a US government export-control directive. The company is part of a growing field of AI developers, including NVIDIA, which is playing a significant role in the development of AI infrastructure.
Key Takeaways
• Most organizations don't treat AI agents as identities, creating security risks.\n• Snowflake's approach to agentic security involves three layers: data, model, and agent.\n• Cisco is investing in quantum security and enterprise AI through partnerships.\n• AI adoption faces gaps despite growing investment, with many organizations struggling to implement AI projects.\n• Perplexity has launched Brain, a self-improving memory system for AI agents.\n• Anthropic took its Claude Fable 5 and Mythos 5 AI models offline due to a US government export-control directive.\n• Generative AI hype may be poisoning the well for traditional AI research.\n• Cybercrime economics are shifting with AI, with threat actors using AI-powered tools for targeted attacks.\n• Monash University has unveiled a secure AI supercomputer for health research.\n• NVIDIA is playing a significant role in the development of AI infrastructure.Most Organizations Don't Treat AI Agents as Identities
Most organizations don't treat AI agents as identities, which creates security risks. AI agents are now part of the enterprise, taking actions and making decisions. However, most security teams don't have proper governance and security measures in place for these agents. This oversight leaves organizations vulnerable to security breaches and data exfiltration.
Shadow AI's Real Threat Is Access Control
The real threat of shadow AI is access control, not data leakage. Shadow AI refers to AI agents created by employees and business units, often without proper oversight. These agents can connect to critical business services, retrieve information, trigger workflows, and update records. This creates a risk profile that's different from traditional shadow IT.
Securing Agentic Enterprise Starts with Data
Securing the agentic enterprise starts with data governance. Agentic AI combines reasoning, data access, and action, making it powerful but difficult to handle. Snowflake's approach organizes agentic security into three layers: data, model, and agent. This framework ensures defense in depth across the full workflow.
AI Literacy Is Everyone's Responsibility
AI literacy is everyone's responsibility in higher education. The key to embedding AI literacy in the workforce of tomorrow is collaboration over competition. A focus on three components of AI education - fundamentals, ethics, and application within a discipline - ensures that students are equipped with the skills they need.
Former AI Head Warns of Generative AI Hype
Generative AI hype may be poisoning the well for traditional AI research. The former AI head at Take-Two warns that excessive hype around generative AI could lead to a backlash against all AI research if the bubble pops.
Cybercrime Economics Shift with AI
Cybercrime economics are shifting with AI trades mass volume for lethal precision. Threat actors use AI-powered tools to bypass multi-factor authentication and launch targeted phishing attacks. Encrypted traffic now hides 95.2% of phishing attempts.
Monash Unveils Secure AI Supercomputer
Monash University unveils a secure AI supercomputer for health research. The MAVERIC system is a trusted research environment for authorized researchers working on sensitive data and AI-enabled biomedical discovery.
Cisco Pushes Further into Quantum Security
Cisco pushes further into quantum security and enterprise AI. The company is a founding partner in the Quantum Communication Fieldlab Rotterdam and launched the Secure AI Factory with Equinix and NVIDIA.
Perplexity Launches Brain, a Self-Improving Memory System
Perplexity launches Brain, a self-improving memory system that builds a context graph of an agent's work and learns overnight. Brain is designed to improve the performance of AI agents over time.
AI Adoption Faces Gaps Despite Growing Investment
AI adoption faces gaps despite growing investment. Many organizations struggle to move AI projects from proof-of-concept to production due to legacy systems, talent shortages, and weak foundations.
Britain's Debt, AI's Burden & Poland's Boom
A look at Britain's debt woes, AI's local impact, Poland's economic boom, and the importance of aluminum. Britain's economic situation is marked by a significant debt problem, while Poland is experiencing a notable economic boom.
Dangerous AI Models Are Coming
Dangerous AI models are coming no matter what. Anthropic took its new Claude Fable 5 and Mythos 5 AI models offline following a US government export-control directive.
Sources
- Every AI Agent Is an Identity. Most Organizations Don't Treat Them That Way
- Forget Data Leakage: Shadow AI's Real Threat Is Access Control
- Securing the Agentic Enterprise Starts with the Data
- ‘AI literacy is everyone’s responsibility’
- Take-Two's ex-AI boss says the generative AI hype is "poisoning the well" and could ward off all use of traditional AI in the future
- Zscaler Research Finds Cybercrime Economics Are Shifting as AI Trades Mass Volume for Lethal Precision
- Monash unveils secure AI supercomputer for health research
- Cisco (CSCO) Pushes Further Into Quantum Security And Enterprise AI
- Perplexity Launches Brain, a Self-Improving Memory System That Builds a Context Graph of an Agent’s Work and Learns Overnight
- CGI: Why AI Adoption Faces Gaps Despite Growing Investment
- Britain's Debt, AI's Burden & Poland's Boom
- "Dangerous" AI models are coming no matter what
Comments
Please log in to post a comment.