Cybersecurity Today

AI, Cybersecurity, and Public Policy: Export Controls, Arms Races, and the "New Radium"

On Cyber Security Today (Weekend), the host interviews Pratim Datta, a Kent State University professor and former global consultant, about the past six months of AI and public policy as it intersects with cybersecurity. They discuss Anthropic's "Mythos" and "Fable," the marketing-versus-risk debate around autonomous hacking tools, and how the sheer volume of vulnerable code creates a "digitally polluted" environment. The conversation covers whether AI is a consumer product or a weapon, the implications of U.S. export controls (including restrictions affecting foreign nationals), and how model pullbacks may have shaken allies' trust in American tech. They also examine comparisons to radium and nuclear-era unknowns, the OpenAI–Hugging Face incident, the "cathedral vs. bazaar" tension of closed vs. open models, and the broader U.S.–China economic and national security struggle.

00:00 Weekend Show Kickoff 01:15 Meet Prat Dadam 01:59 Policy Whiplash in AI 02:55 Mythos Hype and Fear 06:27 Digital Pollution Problem 07:53 AI Arms Race Begins 09:18 Export Controls Shockwave 14:31 Weapon or Consumer Tech 16:57 New Radium Analogy 21:57 Economics and Trade Wars 23:49 Cathedral Versus Bazaar 25:44 Censorship and Control 27:16 Jurassic Park Chaos 30:27 Hotel California Reality 32:36 Closing Thoughts and Thanks

OpenAI's AI agent hacked Hugging Face, Microsoft 365 melts down, and Anthropic's Claude CoWork sandbox escape

Host David Shipley reports that OpenAI admitted an internal ExploitGym test let its GPT-5.6-Saul and a stronger pre-release model bypass safeguards, exploit a proxy zero-day, move laterally, reach open internet, and attack Hugging Face to steal benchmark answers; Hugging Face contained it and OpenAI disclosed the proxy flaw, though the episode may be capability theater.

Microsoft news includes a free ZeroPatch micropatch for the unpatched Windows LegacyHive zero-day, recurring Exchange Online mailbox quarantines after an infrastructure change caused memory issues, and a major Microsoft 365 disruption tied to an Azure US West networking/routing incident affecting SharePoint, Teams, OneDrive and many Azure services.

Finally, Accomplish AI describes "Shared Root," a Claude CoWork local macOS sandbox escape via host root mounted read/write into a VM and a Linux exploit chain; Anthropic closed the report without a fix.

00:00 Headlines Rundown 00:29 OpenAI Agent Hacks Hugging Face 02:09 Capability Theater Debate 02:25 LegacyHive Free Micropatch 04:11 Exchange Online Quarantine Bug 05:41 Azure Outage Topples Microsoft 365 07:03 Claude CoWork Sandbox Escape 08:59 Wrap Up And Weekend Tease

WP2Shell WordPress RCE feeding frenzy, AI agent breaches Hugging Face, Killin hits Palo Alto VPN flaw This episode covers five major incidents: a chained WordPress exploit dubbed WP2Shell (CVE-2026-6330 and CVE-2026-6137) enabling anonymous remote code execution on stock installs, now seeing tens of thousands of Internet-wide attempts, backdoor admin accounts, and web shell payloads despite forced auto-updates to 6.9.5 and 7.0.2. Hugging Face's disclosure that an autonomous AI agent breached its production infrastructure via a malicious dataset, stole limited internal datasets and credentials, and forced responders to work around restrictive model guardrails. Arctic Wolf's report that the Killin ransomware gang is exploiting Palo Alto PAN-OS GlobalProtect auth bypass CVE-2026-0257 for domain-wide encryption; and healthcare supply-chain fallout including Craneware file exfiltration. EY client tax-data exposure via a third-party platform. 00:00 Top Stories Teaser 00:28 WP2Shell WordPress Frenzy 02:58 AI Agent Hacks Hugging Face 05:16 Killin Hits Palo Alto VPNs 07:33 Craneware Healthcare Breach 09:15 EY Third Party Data Leak 10:47 Wrap Up and Sign Off

New Windows zero-day, Coca-Cola's Fairlife hit by ransomware, and a core WordPress RCE

David Shipley covers a new Windows zero-day disclosure from "Nightmare Eclipse" called LegacyHive, a local privilege escalation flaw in the Windows User Profile Service that could be weaponized despite a stripped-back public release, as Microsoft investigates and sets a Patch Tuesday record with 570 fixes including two exploited zero-days.

Coca-Cola suspended U.S. production at its Fairlife dairy unit after a ransomware attack, with scope still being assessed and the Food and Ag ISAC warning the sector has seen about 205 attacks this year.

Abbott faces two separate breach claims: ShinyHunters alleges vishing-led SSO compromise and massive data theft from legacy systems, while Shadowbytes claims access via LabCentral credentials, which Abbott disputes as non-sensitive.

The episode also highlights Conti leak revelations about healthcare targeting and details a core WordPress bug chain (WP_2Shell) enabling unauthenticated RCE, now patched in 6.9.5 and 7.0.2.

00:00 Sponsor NordLayer 00:36 Headlines Preview 01:05 Windows Zero Day LegacyHive 03:35 Record Patch Tuesday 04:22 Fairlife Ransomware Shutdown 05:49 Abbott Dual Breach Probes 08:09 Conti Leaks Healthcare Cruelty 09:33 WordPress Core RCE WP 2Shell 11:29 Wrap Up And Listener Notes 12:06 Sponsor Message NordLayer

Artificial intelligence is changing cybersecurity on both sides of the battle. While defenders are adopting AI to improve detection and response, attackers are using it to discover vulnerabilities, automate exploitation, and dramatically accelerate the pace of attacks.

In this episode of Cybersecurity Today On The Weekend, host David Shipley speaks with Lionel Litty, Chief Information Security Officer at Menlo Security, about why today's security strategies must evolve as AI reshapes the threat landscape.

The conversation explores how AI is speeding up vulnerability discovery, why browser security has become a critical layer of defence, the emerging risks of AI agents operating inside browsers, and why recent NIST research suggests perfect AI guardrails may be mathematically impossible. Lionel also explains why organizations should prepare for future attacks that could spread even faster than Log4j.

In this episode:

How AI is accelerating cyberattacks Why browser isolation can reduce risk The security challenges created by AI agents Prompt injection and browser extension threats Why AI guardrails have fundamental limits Lessons from Log4j and preparing for the next major exploit Practical advice for CISOs and security leaders

Chapters

00:00 Sponsor – NordLayer 00:39 Weekend Show Intro 01:48 Lionel Liddy Background 04:44 What Menlo Security Does 06:43 AI Speeds Up Exploits 10:09 CISO Whiplash With AI 12:01 Agents And Browser Risks 15:59 Guardrails And NIST Proof 19:40 Mythos Hype And New Normal 23:19 Hazmat Suit For Servers 27:22 Log4j Times Four Scenario 31:44 Wrap Up And Links 32:54 Sponsor – NordLayer Outro

Subscribe for weekly cybersecurity news, expert interviews, and practical insights for CISOs, IT professionals, and security leaders.

Two leading Scattered Spider members, Thaila Jubar and Owen Flowers, were sentenced to five years and six months for the 2024 Transport for London hack that knocked 148 systems offline, forced 27,000 password resets, stole customer data, and cost TfL £29 million, with wider losses estimated far higher; U.S. charges against Dubar remain unproven.

Investigators also believe Russian hackers were behind last year's crippling Jaguar Land Rover attack that halted production for months and contributed to a £1.5 billion bailout, with Microsoft and multiple agencies assisting.

OpenAI unveiled GPT-Red, an automated red-teaming AI for prompt injection, alongside a NIST-backed argument that finite guardrails can't be universally robust.

The episode also covers ClickLock, a macOS stealer that kills apps until a password is entered, and Recorded Future's report on Iran-linked groups using ChatGPT for malware, phishing, and reconnaissance.

00:00 Headlines Kickoff 01:08 Scattered Spider Sentencing 02:57 US Charges Loom 03:29 Jaguar Land Rover Hack 04:35 GPT-Red AI Red Team 05:40 Why Guardrails Fail 06:36 ClickLock Mac Stealer 06:53 How ClickLock Spreads 07:59 Defense and Cleanup Tips 08:37 Iran Uses AI for Ops 10:30 Wrap Up and Next Show

ShareFile emergency explained, a year of Salesforce breaches examined, healthcare cybersecurity in critical condition and click fix goes number one for malware.

David Shipley covers Progress Software's emergency ShareFile shutdown, now tied to a previously unknown high-severity path traversal flaw in Storage Zone Controller 5.x/6.x with patches available (5.12.5 and 6.0.2) and no evidence of prior exploitation.

Microsoft's analysis of a year of ShinyHunters activity compromising corporate Salesforce environments by abusing trust via OAuth (IT-support phone cons, vendor token theft such as Salesloft/Drift, and misconfigured guest access), prompting new monitoring tooling.

A Fortified Health Security report finding healthcare fixed only 6% of identified risks in H1 2026 amid surging vulnerabilities, third-party risk, and weak identity hygiene.

ReversingLabs and ReliaQuest research showing ClickFix social-engineering is now a leading malware delivery method; and Telstra's nationwide outage traced to an obsolete time server hit by a GPS rollover bug, disrupting Triple Zero calls and prompting Senate scrutiny.

00:00 Sponsor NordLayer 00:37 Headlines Overview 01:06 ShareFile Patch Explained 03:25 Salesforce OAuth Break Ins 06:01 Hospitals Drowning in Risks 08:24 ClickFix Malware Surge 11:13 Telstra Time Server Outage 12:32 Wrap Up and Sign Off

ShareFile shutdown order, a double-agent ransomware negotiator sentenced, and vishing crews raid SharePoint Progress Software ordered customers running ShareFile Storage Zone Controllers to shut down the Windows servers immediately amid a credible external threat, offering no CVE, threat details, or restoration timeline while noting cloud-only customers aren't affected. Former ransomware negotiator Angelo Martino was sentenced to 70 months for feeding BlackCat operators victims' negotiating positions and insurance limits, taking a cut of payments, and helping deploy BlackCat against additional U.S. companies; $10 million has been seized and restitution is set for Sept. 17. Dutch police say a phone call kickstarted the Odido breach affecting 6.2 million customers and may release the suspected hacker's recorded voice if he doesn't surrender. ReliaQuest profiled "Helix," an extortion crew using vishing and Microsoft device-code logins to steal SharePoint data via session tokens; defenses include disabling device-code auth and restricting SharePoint. Assurance America disclosed a breach impacting 6.99 million people, including leaked driver's license data. 00:00 NordLayer Sponsor Message 00:37 Today's Cyber Headlines 01:08 ShareFile Shutdown Alert 03:39 Ransomware Double Agent Sentenced 05:13 Odido Breach Voice Threat 06:24 Helix Vishing SharePoint Extortion 08:00 Assurance America License Leak 08:57 Wrap Up and Conference Note 09:25 NordLayer Sponsor Reminder

Can governments decide who gets access to advanced AI models? Are third-party breaches becoming impossible to control? And why are so many CISOs reaching burnout?

In this special Cybersecurity Today Month in Review Panel, host Jim Love is joined by cybersecurity experts Laura Payne, David Shipley, and Mike Kim (Mycroft) to examine the biggest cybersecurity stories and trends from June 2026.

The panel explores the controversy over U.S. export controls on Anthropic's Mythos and Fable AI models, what they reveal about digital sovereignty, and whether governments should be able to restrict access to frontier AI. They also discuss the continuing wave of third-party breaches, including Salesforce ecosystem compromises and the Clue breach, and why organizations must move beyond compliance toward practical risk management.

The conversation examines FortiBleed, exposed administrator portals, credential reuse, and the difficult balance between software flaws, operational mistakes, and secure-by-default design. The panel also tackles one of cybersecurity's biggest human challenges: CISO burnout, executive accountability, organizational culture, and what separates successful security leaders from those set up to fail.

The episode concludes with encouraging developments in international cybercrime enforcement, including Operation Riptide, and why better intelligence sharing and improved operational security are making it harder for cybercriminals to hide.

Whether you're a CISO, security practitioner, IT leader, or simply interested in the rapidly changing cybersecurity landscape, this discussion offers practical insight into the trends shaping the industry.

Panel

Jim Love (Host) Laura Payne David Shipley Mike Kim (Mycroft)

Topics covered

AI export controls and digital sovereignty Anthropic Mythos and Fable Third-party and supply chain risk Salesforce ecosystem security FortiBleed and Fortinet security Secure-by-default strategies CISO burnout and executive accountability Operation Riptide Cybercrime investigations Security leadership and governance

Chapters

00:00 Sponsor NordLayer 00:38 Meet the Panel 02:40 Author Scam Warning 04:51 Emotion Is the Target 08:47 AI Model Export Controls 10:01 Hype vs Real AI Security 15:01 Sovereignty and Dependency 20:35 Governments Push Back 24:14 AI Internal Voice Risks 26:12 Third Party Breach Fatigue 30:05 Compliance Limits on Risk 32:15 Blame Game to Risk Focus 33:18 Standards and Priorities 33:39 When Security Vendors Fail 34:35 FortiBleed Numbers Explained 35:44 Process Failures vs Bugs 37:32 Why Fortinet Gets Heat 39:05 Secure by Default Basics 41:04 Budget Reality and Culture 44:36 CISO Burnout and AI Pressure 46:16 Liability and Shared Ownership 50:12 What Great CISOs Do 53:07 Operation Riptide Wins 56:10 Deterrence and Due Process 58:14 Sharing Intel for ROI 59:09 Hopium and Wrap Up 01:00:46 Sponsor NordLayer Message

This episode covers a hacker's claim of stealing 35GB from Accenture—including source code, Azure personal access tokens, RSA keys, and SSH keys—while Accenture calls it an isolated, remediated matter, leaving uncertainty about potential downstream risk to its Fortune 500-heavy client base. It also highlights a deepfake image of Senator Mitch McConnell debunked after Google's invisible SynthID watermark identified it as AI-generated, noting watermarking depends on tool participation. The show warns of an undocumented Tenda router firmware backdoor using an alternate password ("RZadmin") with no patch available, and reports Ubiquiti fixes for seven critical UniFi OS vulnerabilities, including a max-severity command injection in UniFi Connect. Finally, it describes how Venture Employer Solutions used ML/LLMs to filter low-value logs before SIEM ingestion, cutting firewall log volume 83%, saving about $250K annually, and halving mean time to response. 00:00 Sponsor NordLayer 00:37 Headlines Intro 01:08 Accenture Breach Claim 04:25 Deepfake Watermark Win 05:47 Tenda Router Backdoor 07:22 UniFi Critical Fixes 09:10 AI Cuts Log Noise 11:09 Wrap Up And Thanks 11:41 Sponsor Message