AI Will Save Us, or Not? - PSW #945

In the security news this week: Build your own router, or just buy one What to patch first But maybe don't buy D-Link AI nearly starts a war Flock cameras lose their keys The AI slowdown won’t save bad security Opus at home, just slower Black Hat says fundamentals still work Hacker gadgets, and my top pick Gyazo screenshots Fake job interviews, real malware VINCE gets a new home Munich university gets disconnected LinkedIn fights the scraping machine SolarWinds hard-codes another bad idea Fake LastPass kills 145 security tools Colorado water gets its settings changed AI CEOs sell apocalypse and bonds The AI safety cult gets audited Ransomware recovery takes weeks, not hours TeamPCP’s supply-chain tour continues Zuckoff hunts smart glasses Visit https://www.securityweekly.com/psw for all the latest episodes! Show Notes: https://securityweekly.com/psw-945

As businesses race to embrace AI, security leaders are struggling to modernize cyber hygiene and prevent over-privileged agents from causing unintended harm. How do you balance the benefits of AI with the Risks of AI? Evan McHenry, Chief Information Security Officer at Robinhood Markets, joins Business Security Weekly to discuss how to practically implement and secure AI in the enterprise. Evan will share his lessons learned over the last 18 months, including how to communicate with the Board, why you should own Enterprise Architecture and embrace IT, and why you don't have time to argue with your CFO. If you're struggling to balance the benefits of AI with the Risks of AI, this interview is for you. In the leadership and communications segment, Security spending is growing — except for the typical CISO, What Leaders Need to Know About AI and Psychological Safety, The Cyber Gap, and more! Visit https://www.securityweekly.com/bsw for all the latest episodes! Show Notes: https://securityweekly.com/bsw-466

WWIII,Security Debt, JFK, CISA, SUSE, OpenAI, Google, DORA, Aaran Leyland, and More on the Security Weekly News. Visit https://www.securityweekly.com/swn for all the latest episodes! Show Notes: https://securityweekly.com/swn-618

Prompt injection demonstrates one of the major challenges in securing LLMs and agents -- how do you ensure an agent ignores attackers and only does what you instructed it to do. The flaw highlights how LLMs mix inputs, context, and outputs without any strict boundaries between them. Julie Brunias joins us to talk through examples of injections, why their consequences can go beyond information leaks, and why trying to mitigate them with other LLMs is insufficient. Segment resources: https://llmgateway.io/open-source https://www.wiz.io/blog/off-guard-breaking-litellm-from-authentication-bypass-to-cloud-compromise Visit https://www.securityweekly.com/asw for all the latest episodes! Show Notes: https://securityweekly.com/asw-401

Interview with Rob Sadowsky from Cohesity Global Cyber Resilience Report: Cyber Recovery Plans Weren't Designed for this Moment Cyber recovery plans weren't designed for this moment. Most were built around assumptions that made sense when they were written: incidents could be understood, dependencies mapped, recovery could follow a predictable sequence, and decision-makers would have enough information to act. In practice, major cyber incidents unravel those assumptions. AI and autonomous agents are creating new paths to compromise, while cloud and SaaS expansion increases the systems, services, and dependencies involved in recovery. Vulnerabilities are discovered and weaponized faster than ever, and AI is accelerating that cycle. As incidents unfold, scope expands, dependencies appear only when they break, and recovery plans no longer match reality. With assumptions under greater strain, confidence is beginning to erode. This year, the percentage of survey respondents reporting complete confidence in their cyber resilience strategy fell. To understand how recovery unfolds today, Cohesity commissioned Vanson Bourne to survey 3,200 IT and security decision-makers at organizations with 1,000 or more employees across 11 countries. This report examines where recovery becomes more difficult than expected, the obstacles organizations encounter, how they define and test a Minimum Viable Company (MVC), and how AI is reshaping cyber threats and cyber resilience. https://www.cohesity.com/dm/global-cyber-resilience-report/ This segment is sponsored by Cohesity. Visit https://securityweekly.com/cohesity to learn more about them! Topic: When should we use AI for writing and when should we avoid it? I've had an essay in draft form for a month now, trying to get my feelings across on why AI writing drives me so crazy. I struggled to put it into words. Fortunately, Charity Majors figured out how to put it into words and I think she nailed not just how I feel about AI, but the reasons why I feel so strongly about it. She uses a scale to help explain this, with "personal" at one end and "functional" at the other. https://charity.wtf/p/confessions-of-an-unrepentant-slop When I ask AI to create a company profile for me, 10 minutes before I meet with them, I don't need poetry - just facts. But when I read something that is supposedly someone's opinions and analysis on a topic, and it's clearly 100% AI-generated, I angrily dismiss it. I love that this writeup isn't just an "I hate slop" rant - it actually quantifies why a personal touch matters and how to gauge when it is necessary and when outsourcing the task to AI is totally fine. In both cases, Charity notes that quality matters. My most recent complaints come from cases where things are not only clearly written by AI, but where quality went out the window and they're unrecognizable as a human-readable language. And yes, I brought an example: https://dispatch.cybersecurityhq.com/p/escalation-voided-on-construct-failure-timing-condition-placed-under-review Weekly Enterprise News Finally, in the enterprise security news, We check the vibes, funding, and acquisitions Tenable now has Mythos built-in??? We check in on how vulnerability remediation is going Microsoft is creating a code of conduct for AI OpenAI just got called to the principal’s office Booz Allen created new cybersecurity AI benchmarks 50% of CISOs see Mythos as a sign to resign??? Ayman read the latest Anthropic AI misuse report MIT explains the 12 possible AI outcomes (very ominous) a 9-year old decided to promote his YouTube account… with his dad’s corporate card All that and more, on this episode of Enterprise Security Weekly. Visit https://www.securityweekly.com/esw for all the latest episodes! Show Notes: https://securityweekly.com/esw-477

Bacteria, Spartans Invade Athens, Agentic AI gone wild, Cisco, WordPress, Settra, Plugin4Shell, Josh Marpet, and More on this episode of the Security Weekly News. Visit https://www.securityweekly.com/swn for all the latest episodes! Show Notes: https://securityweekly.com/swn-617

In the security news this week: UK government rolls out passkeys to 20 million users Phishing-resistant authentication and replay resistance Passkey adoption, device security, and user acceptance EU Cyber Resilience Act guidance, scope, and compliance CRA vulnerability disclosure and reporting requirements The real cost of cyberattacks and cybersecurity spending Cyber insurance and improving organizational security Nightmare Eclipse and the release of Windows zero-days Check Point VPN vulnerabilities and perimeter security GitLab security updates and shadow IT Discovering unmanaged GitLab instances Cyberattacks against oil tankers and insider threats VPN patching and implied rules Zero-downtime GitLab updates and version management Running Windows ARM on Apple Silicon with VMware and Parallels Visit https://www.securityweekly.com/psw for all the latest episodes! Show Notes: https://securityweekly.com/psw-944

The overwhelming majority of people, across the full span of their professional lives, operate without formal authority over the domains in which they work (i.e., leadership). Yet followership has almost no sustained literature, no targeted development, and no rigorous framework of its own. If you're not a leader, what does a follower look like? Kenyada Meadows, CEO & Founder at The Executive Parent Company, joins Business Security Weekly to discuss his book, The PASSENGER Seat, which outlines a framework for followership. Kenyada will outline the nine disciplines of the framework across three dimensions, including: (P) Principled Anchoring — Holds values explicitly and specifically enough to guide behavior under pressure; knows the limit line before reaching it. (A) Accountability — Communicates upward honestly, documents concerns before decisions are made, and owns errors without deflection. (S) Self-Mastery — Invests deliberately in work, home, and inner domains, so identity — and ethical agency — isn't hostage to institutional threat. (S) Systems Thinking — Reads the institution analytically — what produces the patterns observed, and where honest information is being filtered. (E) Emerging Risk Management — Functions as an early-warning system for risks that formal frameworks haven't yet named. (N) Negotiation — Translates insight and values into institutional impact through credibility, framing, coalition, and timing. -(G) Guardianship — Protects what the institution stands for over time, especially when no one is watching. (E) Enablement — Ensures influence strengthens collective capacity rather than merely advancing individual position. (R) Results-Driven Influence — Binds the other eight disciplines toward outcomes that are genuinely better — not merely visible. In the leadership and communications segment, Stop playing with the CISO role. Fix cybersecurity leadership, What CIO-CISO alignment looks like when it's working, Why judgment is emerging as cybersecurity’s defining skill, and more! Visit https://www.securityweekly.com/bsw for all the latest episodes! Show Notes: https://securityweekly.com/bsw-465

RoboGators, HBOMAX, Microsoft, DAS, Horsebot 3000, Aaran Leyland does AI, and More on the Security Weekly News. Visit https://www.securityweekly.com/swn for all the latest episodes! Show Notes: https://securityweekly.com/swn-616

While agents and LLMs haven't fundamentally changed core mobile vulnerability types, they have supercharged speed, scale, and accessibility—democratizing threats like automated phishing, synthetic identity fraud, and easier identification of hard-coded secrets. Ryan Lloyd and Jason Cortlund break down how threat actors leverage LLMs as a force multiplier to accelerate mobile app attacks. Then we discuss actionable defense strategies, from viewing agents as an active adversary to leveraging server-side threat telemetry, attestation, and layered defense strategies combined with polymorphic code releases. This segment is sponsored by Guardsquare. Visit https://securityweekly.com/guardsquare to learn more about them! Visit https://www.securityweekly.com/asw for all the latest episodes! Show Notes: https://securityweekly.com/asw-400