AI safety tests reach the internet, Private Relay flaw unmasks IPs, weak telcos invite Salt Typhoon

OpenAI slows release of Astra model citing cyber capabilities Irregular won't say if there were more rogue incidents U.S. cyber ambassador nominee Cassady confirmed in Senate Episode shownotes: https://cisoseries.com/cybersecurity-news-openai-slows-astra-irregular-wont-talk-senate-confirms-cassady/ Huge thanks to our sponsor, ThreatLocker AI is helping attackers research targets, create malicious code, and adapt faster. But the fundamentals have not changed. Code still needs to execute, applications still need access, and attackers still need privileges. Today's tip: control those actions instead of trying to predict every threat. Learn more from ThreatLocker at threatlocker.com/ciso. 

Apple's bug bounty program overwhelmed by fake AI generated reports China launches cybersecurity review into Palo Alto Networks products Meta AI hacks external systems during cybersecurity testing Get the show notes here: https://cisoseries.com/cybersecurity-news-faked-bug-reports-metas-rogue-ai-china-investigates-palo-alto/ Huge thanks to our episode sponsor, ThreatLocker AI risk does not only come from attackers. Employees are adopting AI tools faster than many organizations can evaluate them. Today's tip: an AI policy should be backed by enforceable controls over what tools can access and do. See how ThreatLocker can help you govern AI use at threatlocker.com/ciso.

AI safety tests spill onto the real internet Private Relay flaw unmasks IP addresses Weak telcos left door open for Salt Typhoon Get the show notes here: https://cisoseries.com/cybersecurity-news-ai-safety-tests-reach-the-internet-private-relay-flaw-unmasks-ips-weak-telcos-invite-salt-typhoon/ Huge thanks to our episode sponsor, ThreatLocker AI is compressing the timeline between initial access and impact. That leaves security teams less time to identify, investigate, and contain malicious activity. Faster response helps, but prevention can remove actions from the attack chain entirely. Give your responders a more controlled environment. Book a ThreatLocker demo at threatlocker.com/ciso.  

ChainDrop attack hits npm Pass-ta-key attacks target passkeys  AI accounts for most cybercrime in Africa Get the show notes here: https://cisoseries.com/cybersecurity-news-chaindrop-hits-npm-pass-ta-key-targets-passkeys-ai-runs-amok-in-africa/  Huge thanks to our episode sponsor, ThreatLocker Attackers do not always bring their own tools. AI can help them find ways to misuse software already trusted by the organization. Today's tip: approval should not mean unlimited access. Define what trusted applications can reach and what they can do. Learn how ThreatLocker applies this principle at threatlocker.com/ciso today.  

When license plate readers become stalking tools ExfilSquad dumps UK police contact data China-linked hackers shrink the patch window Get the show notes here: https://cisoseries.com/cybersecurity-news-license-plate-readers-as-stalking-tools-exfilsquad-dumps-uk-police-data-the-ever-shrinking-patch-window/ Huge thanks to our episode sponsor, ThreatLocker An attacker uses AI to create a payload your security tools have never seen. Detection now has to recognize it before the payload can act. ThreatLocker approaches the problem earlier by controlling whether that code is permitted to run at all. Explore a deny by default approach for your organization at threatlocker.com/ciso.  

UK's state investments agency suffers data breach CISA tells utilities to remove internet-exposed PLCs following Minnesota attacks Anthropic says its AI hacked real-world companies in three incidents Get the show notes here: https://cisoseries.com/cybersecurity-news-cybersecurity-news-uk-investments-agency-breach-cisa-water-warning-anthropic-threesome/ Huge thanks to our episode sponsor, ThreatLocker AI is helping attackers research targets, create malicious code, and adapt faster. But the fundamentals have not changed. Code still needs to execute, applications still need access, and attackers still need privileges. Today's tip: control those actions instead of trying to predict every threat. Learn more from ThreatLocker at threatlocker.com/ciso.  

This week's Department of Know is hosted by Rich Stroffolino, with guests Janet Heins, CISO, ChenMed, and Derek Fisher, Director of the Cyber Defense and Information Assurance Program, Temple University. Missed the live show? Check it out on YouTube. The Department of Know is live every Friday at 4:00 p.m. ET. Join us each week by registering for the open discussion at CISOSeries.com. Huge thanks to our sponsor, Pindrop A finance worker joined a video call with their CFO and wired $25 million to attackers. This isn't fiction—it happened. Deepfake video. AI voice. Completely convincing. It could be happening in your meetings right now. Pindrop Pulse for Meetings can detect deepfake impersonation before the damage is done. Go to pindrop.com and start verifying.

Semiconductor firm Analog Devices discloses data breach Copilot for Word POC copies hidden prompts into new documents Microsoft Teams vishing attacks lead to Chaos ransomware attacks Get the show notes here: https://cisoseries.com/cybersecurity-news-analog-devices-breach-copilot-ai-worm-teams-ransomware-vishing/ Huge thanks to our sponsor, Pindrop A finance worker joined a video call with their CFO and wired $25 million to attackers.   This isn't fiction—it happened. Deepfake video. AI voice. Completely convincing.   It could be happening in your meetings right now. Pindrop Pulse for Meetings can detect deepfake impersonation before the damage is done. Go to pindrop.com and start verifying.  

OpenAI agent's escape reaches a Modal customer Hackers hit Minnesota water systems Fake Russian companies, real stolen money Get the show notes here: https://cisoseries.com/cybersecurity-news-openai-agent-reaches-modal-minnesota-water-systems-hacked-fake-russian-companies-steal-real-money/ Huge thanks to our sponsor, Pindrop TIME named Pindrop one of the 10 Most Influential Software Companies of 2026.   Deepfakes slip into your video meetings, contact centers, and hiring pipelines. Pindrop restores trust to every digital conversation. Customers. Employees. Defended.   Don't wait for an incident report. Visit pindrop.com.  

Thousands of server BMCs leak password hashes Claude finds flaws in encryption Google gives old threat actors new names Get the show notes here: https://cisoseries.com/cybersecurity-news-leaky-bmcs-claude-finds-encryption-flaws-googles-new-names/  Huge thanks to our sponsor, Pindrop AI attacks on the enterprise are skyrocketing. Is your tech stack keeping up?   Deepfake and synthetic voices are slipping through a channel your defenses were never built to cover—stealing credentials and exposing data with no visibility until after the incident report.   Pindrop closes that gap, with under 1% false positives. Go to pindrop.com.