Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

Enterprise RAG pipelines have a recall stage and a precision stage. The retriever handles recall. The reranker handles precision. Skipping the reranker, or misplacing security controls around it, is where most accuracy and data exposure problems begin. The retriever’s job is to pull back every document that might be relevant. The reranker’s job is to find, from that candidate set, the documents that actually answer the question.

Most agentic triage solutions hand you a verdict and stop right there. Maybe you get an alert ranking, maybe a severity score, maybe some light investigation, but you, as the analyst, still have to do the cleanup manually. There is still a mountain of evidence that needs to be gathered by hand, and a mountain of cases to dig through on your own.

Security teams didn’t need a headline to tell them that vulnerability volumes continue to be problematic. The CVE database now contains over 354,000 records. Annual disclosure rates have climbed steadily for more than a decade. And remediation backlogs have long been recognized not as an aberration, but as a fixture of the job.

A software bill of materials works because software changes through a build. Someone bumps a dependency, the pipeline runs, the manifest updates and a scanner compares the new list against known vulnerabilities. Every part of that loop assumes a rebuild is the thing that changes behavior. ‍ AI systems break that assumption at the point it matters most. Editing a system prompt changes what a model does, swaps no dependency, triggers no build and produces no new manifest.

‍ The distance between what OpenTelemetry was built for and what AI governance is asking of it shows up in a single number. Distributed tracing descends from Dapper, the 2010 Google paper that gave the industry the vocabulary of traces and spans. Dapper sampled one trace in 1,024. That is ample for finding a latency regression, because a regression recurs and the next sample catches it.

Every CISO managing an API estate has heard of Broken Object Level Authorization (BOLA) and Broken Function Level Authorization (BFLA). What is harder to pin down is what these vulnerabilities actually cost the business when they go unaddressed. Board members and finance teams want numbers, not acronyms, and that gap between technical risk and financial risk is where security budgets get lost. BOLA has held the number one spot in the OWASP API Security Top 10 since the list was created in 2019.

Claude Mythos is arguably the strongest cybersecurity model that Anthropic has built. But we know that model capability is only part of what determines how well an AI vulnerability product performs. To test that, we put Anthropic’s Claude Security, which runs on Mythos, and Aikido Code Security Audit head-to-head on the exact same target to see which harness can deliver the best coverage and at what cost. Code Security Audit is part of Aikido’s AI Code Analysis suite.

The riskiest thing an AI agent does all day isn’t writing code. It’s shopping. Every few minutes, it reaches out for a package, an AI asset, or a tool, and pulls it in with no real way to check what’s inside. We think the fix is agent immunization: security that lives inside what an agent consumes, builds, and ships, not a wall built around it.

For twenty years, the software security playbook has worked the same way. You find the vulnerability, score it, open a ticket, assign it to a human, wait for the fix, ship the patch, and prove it happened. Every step in that sequence assumes humans can review each fix individually and still keep up. Frontier AI broke that assumption. The exploit window has collapsed from weeks to hours. Attackers reason across your codebase, chain their findings, and ship exploits before a CVE is even published.

AI is making social engineering attacks significantly more effective, according to a new report from cyber insurance firm Resilience. These attacks were behind more than 85% of losses in the first half of 2026, compared to less than 20% during H1 2024. “Losses tied to phishing, social engineering, and transfer fraud have climbed from 17.7% of incurred losses in H1 2024 to 85.3% in H1 2026, the single largest increase in the report’s five half-year comparison,” Resilience says.