Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

Last Tuesday, October 21st, OpenAI released ChatGPT Atlas, an AI-powered browser that allows users to interact with ChatGPT directly from any browser tab. Throughout last week, the Cyberhaven Labs team tracked its adoption in corporate environments and actively investigated its security vulnerabilities.

AI is rewriting the rules of privileged access, but the rise of AI agents is creating a governance crisis. Threats like credential stuffing and privilege escalation are now accelerated by autonomous systems moving faster than humans can react. 82% of companies deploy autonomous AI agents, but 23% of IT teams admit those bots have already been tricked into revealing credentials—and fewer than half have guardrails in place. In modern infrastructure, machine identities now outnumber humans 80:1.

At the AI Agent Security Summit in San Francisco, some of the brightest minds in AI security and top industry leaders gathered to tackle one of the most challenging problems in tech nowadays - how do we secure super smart systems that change at runtime and are designed to think, adapt, and compete? As someone who spends every day turning AI security challenges into tangible solutions, I left the summit both inspired by the innovation on display and concerned by the magnitude of what’s still ahead.

Static application security testing (SAST) is foundational to modern application and code security programs. Yet these tools inevitably produce false positives that require manual review. When scanners find vulnerabilities that are not genuine issues, they erode trust, slow down remediation, and make it harder for teams to understand which alerts require attention.

Artificial intelligence has only been available for a relatively short period. Still, already many cyber defenders are as frightened as if Jenna Ortega’s Wednesday Addams had whipped her head around and set her dark, dangerous eyes on them. It’s not hard to see why. Machine learning, Gen AI, and Retrieval-Augmented Generation (RAG) are a few of more than 20 new acronyms flooding our industry, with more being added almost every day.

I’ve been spending a lot of time with teams and customers talking about AI. Not in terms of buzzwords or market predictions, but the real, in-the-trenches work of building software, serving customers, and securing identities and data. The mindset we’ve adopted around AI is simple: you can’t cut your way to great products or great customer experiences. AI isn’t about replacing people or chasing short-term efficiency gains.

This piece is part of a monthly series by Carisa Brockman and Bindu Sundaresan exploring the evolving world of AI governance, trust, and responsibility. Each month, we look at how organizations can use artificial intelligence safely, thoughtfully, and with lasting impact.

AI convenience rides on a river of data: text, clicks, images, voices, locations, and metadata you didn’t know existed. The core question is not whether AI uses data but how it collects it, what it infers, and whether people truly agree to that. In other words, the impact of AI on user consent and data collection is not academic. It decides whether your product earns trust or burns it.

It’s rare to find a security professional who enjoys the security information and event management (SIEM) migration process. Security teams are often diverted from actively defending their environment to tend to the long, manual, and error-prone migration process.