AI attacks are now autonomous, adaptive, and multi-step. Static prompts won't cut it. TrojAI Detect's agent-led red teaming simulates multi-turn attacks, adapts to responses, and maps findings to OWASP, MITRE & NIST. Read more: bit.ly/4k120rn

As MCP connects agents to tools and data, risk shifts to the action sequences they follow. Prompt injection, tool poisoning, and multi-step data exposure need new defenses. Traditional AppSec isn't enough. Govern the agent loop. Learn more: bit.ly/4Bpeob2

Bots, fraud, and AI-powered attacks are growing more sophisticated in retail, and PCI-DSS compliance alone doesn't address them. Watch this on-demand webinar for smarter attack prevention strategies for retail security teams. ▶️ bit.ly/4eS77WS

A10 joined 100+ attendees at the Infosecurity Finance Sector Seminar in Taipei, where A10's Nick Chen discussed preventing and mitigating threats in AI-powered environments. Thanks to everyone who connected with us! 🤝 What’s your team’s top AI security priority?

Prompt injection has no fixed signature. Jailbreaks unfold across turns that each look harmless in isolation. And AI risk is bidirectional. Catching AI threats requires detection that reasons about intent across the full request. Learn more. ➡️ bit.ly/4wRVt58

OWASP now maintains two AI security frameworks. The LLM Top 10 protects how the system thinks. The Agentic Top 10 protects what it can do once output becomes action. Modern AI systems both think and act. The security strategy has to match. ↪️ bit.ly/4wqqTPJ

Financial services firms lead all industries in AI adoption. Yet only 16% say their infrastructure meets AI demands with capacity to spare. 📊 A10’s State of AI Infrastructure Report breaks down what modernization needs to look like: bit.ly/4bI4fLy

The first generation of AI gateways forwarded requests. But is your AI gateway modernized? The A10 AI Gateway matches each request to the right model, for the right user, at the right cost. Learn more: bit.ly/45cscXz

Safety-aligned AI models resist helping test their own defenses. Learn how TrojAI by A10 Networks uses model abilteration to keep safety evaluation ahead of frontier models. ➡️ bit.ly/4yS8cWO

One subnet absorbed 11 DDoS amplification vectors over 19 days, cycling through DNS, NTP, CLDAP. 🎯 And protocols from 1983 that most defenders consider obsolete. Attackers rotate until one gets through. Single-vector mitigation stops one channel. ➡️ bit.ly/4oPNmDU

Is your AI red-team assessment already outdated? A one-time test can't keep up with changing models, tools, and attackers. Join Jamison Utter and Priyanka Mullan for a LinkedIn Live on Oct 28 at 9 AM PT. Register now: bit.ly/3WooVmB

A10 Networks is sponsoring GPSEC in Fort Lauderdale on Oct. 28, hosted by @GuidePointSec.🌴 Come find Will Quinto and Vyoma Gandhi to discuss application and AI security strategies! Request to register: bit.ly/46ZfZ9A

A10's Paul Kenniston will join a panel on AI and emerging tech at the Cybersecurity Summit in Houston to discuss exactly how enterprises can safely navigate this shifting landscape. Use discount code CSS26-A10 and register at the link: cyberriskalliance.swoogo.com…

A10 Networks will be at CyberVerse 2026 in London on October 22, hosted by Infinigate UK. 🎡 AI is front and center, and we’re looking forward to the conversation about delivering secure, scalable AI applications. Register. 👉 bit.ly/471Q15i

When agentic AI generates its own APIs, the exposure risks they create require a fundamentally different security approach. A10 security experts explain why SecOps teams need specialized visibility and governance now. ⬇️ youtube.com/shorts/kQiCYUsy_…

Some of the most damaging API breaches in recent memory trace directly back to BOLA, the #1 API security risk in the OWASP API Security Top 10. A10's Carlo Alpuerto and Randy Franklin Smith break down how it works and how to stop it. Watch on demand. ➡️ bit.ly/429nmss

Once AI can call tools via MCP, an ordinary request like “summarize customer data” can pull records the requester shouldn’t see. Inspecting prompts alone won’t catch it. Check intent before the tool call and data in the response after it. Read more ➡️ bit.ly/4wRVt58

49% of respondents identify security as their biggest limitation with AI We built a tool to help. Schedule a live demo today. bit.ly/41XQl2B

Fake referral accounts, thousands of automated signups, and credits funneled to one master account. That's business-logic fraud at scale, but standard defenses often miss it. A10's Gary Wang shares two strategies to build a layered defense. youtube.com/watch?v=XxRMx6N3…

Every AI deployment has two systems: a deterministic layer to harden and a probabilistic layer to contain. Treating them as one is where AI security breaks down. You don’t secure a model in isolation. You secure the system that embeds it. Here's how. ➡️ bit.ly/4hZ0ntb