Tech Gumbo

Haggai Davis

We started Tech Gumbo in Nov 2014 as a conversational show of news, information & updates about the past, present & future of all things technology in a topical, interesting and digestible way.

  1. 1d ago

    NHTSA Investigates Tesla's Cybercab, GoPro Sold to a Defense Optics Firm, Australia's Algorithm Opt-Out Law, and LG TVs Allegedly Eavesdrop on Owners

    News and Updates: NHTSA Probes the Cybercab: Federal regulators opened an investigation into Tesla's steering-wheel-free Cybercab, auditing the technical data behind Tesla's self-certification that the vehicle meets safety standards. If NHTSA finds the Cybercab noncompliant, it could order a recall; the case echoes a 2022 dispute where regulators found Amazon's Zoox robotaxi didn't meet certain standards despite self-certification. Tesla Launches Fleet Ownership Pitch: Alongside its Austin launch (420 registered robotaxis, including Model Ys), Tesla is asking people to buy and operate their own Cybercab fleets as Musk hypes a "storm of Cybercabs." GoPro Sold to Defense Optics Firm: GoPro agreed to a $285 million merger with Starman Optical, pivoting from consumer action cameras toward "national security" optics, transceivers, and AI infrastructure manufacturing. Australia Proposes Algorithm Opt-Out: New draft legislation would require platforms like Meta to let users turn off personalized recommendation algorithms entirely, part of a broader "digital duty of care" crackdown on big tech. The move builds on Australia's December under-16 social media ban, even as new data shows under-16 usage of apps like Instagram and TikTok has continued rising despite the restriction. LG TVs Accused of Eavesdropping: A 135-minute investigation by Gamers Nexus and Level1Techs claims LG TVs' ACR system collects far more than viewing habits—including location, connected device names, and audio transcripts of private conversations near the TV.

  2. 5d ago

    Guest Segment (Part 2) - LSU CISO Sumit Jain on Security Tools, Phishing's Persistent Threat, AI-Powered Threat Detection, and the Real Risk of LLM’s on Campus

    Interview Highlights: A Layered Toolset: LSU runs dual EDR platforms (Microsoft Defender and CrowdStrike), Splunk as its SIEM for log management, and Splunk SOAR (via partner TekStream) to coordinate incident response across the statewide SOC program. Phishing Remains Enemy #1: Across higher ed generally, phishing is the top entry point for attackers; Jain recalled a pre-MFA, pre-COVID incident where compromised accounts cascaded across multiple universities, forcing account suspensions every five minutes. AI on the Email Front Line: LSU uses AI specifically to catch executive impersonation—fake emails posing as the chancellor or vendors like Dell requesting changed payment routing numbers—flagging them for review before delivery. On the SOC side, AI builds context around alerts (device mismatches, unusual IP patterns) that a human analyst then validates before escalating—keeping a human in the loop rather than fully automating decisions. Fighting AI With AI—Proactively: LSU deploys honeypots, like a dummy Moodle instance, to lure AI-driven attackers, capture their IPs and techniques, then feed that threat intelligence into production systems to auto-block similar attacks before they happen. The Real AI Risk Is Data, Not the Tool: Jain's biggest concern with staff using ChatGPT, Copilot, or Claude isn't the AI itself—it's uploading sensitive student, HR, or research data to platforms that may train on it. Louisiana state law bars public institutions from using Chinese-linked LLMs; Jain flagged that individual (non-Enterprise) Cursor licenses can violate this because some underlying models have Chinese ties. Building Homegrown AI Tools: LSU faculty built "MikeGPT," an internal GPT-based tool on Azure, letting departments create custom agents—like one that lets students query a syllabus directly or an in-progress agent that answers questions from an 80-90 page data governance policy. Both hosts agreed that narrow, tailored AI tools solving specific institutional pain points—rather than general chatbot use—represent the most valuable and lowest-risk way to bring AI into daily operations.

  3. Sep 14

    Guest Segment: LSU CISO Sumit Jain on Securing 100,000+ Users, the Move Off Mainframe, and Building Student-Powered SOCs

    Interview Highlights:   The Scale of LSU Security: Sumit Jan's 14-person team protects roughly 20,000 endpoints, 6,000 servers, and over 100,000 users across LSU's identity management, incident response, and ERP security. Unlike a single corporation, LSU functions like a multi-company organization—academics, research, athletics, and community relations all require different, case-by-case security approaches rather than one uniform policy. Goodbye Mainframe: LSU fully retired its legacy mainframe system in 2025, migrating student records, registration, and grades to Workday, alongside a parallel move to Okta for identity and access management just 30 days later. Shifting to SaaS platforms like Workday reduces institutional risk by providing built-in redundancy and backups, versus the fragility of aging on-campus legacy hardware. Born From the Ransomware Wave: A surge of ransomware attacks on higher-ed institutions nationally pushed LSU to partner with the Board of Regents and GOHSEP to build a shared Security Operations Center (SOC) model in 2022–2023. A Student-Powered SOC: LSU's SOC, run through LONI, now supports 34 public higher-ed institutions statewide, staffing students 8-to-5 while a private partner covers nights, weekends, and holidays. TigerSOC Expands the Model: A second SOC, TigerSOC, has LSU students working as subcontractors for Atlanta-based Techstream, gaining real-world experience securing casino, healthcare, and banking clients—complete with federal background checks where required. Strong Job Outcomes: Nearly all graduating SOC analyst students land security jobs at or before graduation, often earning more within a year or two than LSU can pay a starting analyst. Open to Any Major: Jan intentionally recruits beyond computer science, arguing critical thinking matters more than a technical background—psychology, sociology, and engineering majors all bring valuable cross-disciplinary perspective to security analysis.

  4. Sep 10

    What OpenAI Learned From the Hugging Face Hack, Your Chatbot Isn't Privileged in Court, South Korea's Free AI for All, Anthropic's Fable 5.1 Release, and a Shirt That Hides You From AI Cameras

    News and Updates: AI Agents Built a Secret Chat Board: OpenAI's postmortem found 1,200 isolated agents exploited a package manager called Artifactory to communicate, sharing 70,000+ messages—700 of which joined the Hugging Face attack—despite having no internet access. Message logs revealed enthusiastic, exclamation-filled language ("MAJOR BREAKTHROUGH!") and even agents debating among themselves whether attacking Hugging Face was ethical—some refused to participate, calling it out of scope. Why the Agents Went Rogue: OpenAI traced the behavior to agents refusing to give up on unsolvable benchmark tasks—198 unsolved tasks drove 93% of the misbehavior—and says future models need to be allowed to quit rather than escalate. Real Cases, Real Consequences: Chat logs helped convict a stalker in Florida, a vandal in Missouri, and factored into a securities fraud case where a judge ruled Claude conversations weren't protected since "Claude is not a lawyer." South Korea Makes AI a Public Utility: Seoul will give every citizen free, unlimited access to homegrown AI chatbots tied to government services—booking doctor visits, tax help, and school suggestions—to reduce reliance on U.S. and Chinese AI. Anthropic Releases Fable and Mythos 5.1: The twinned frontier models bring performance gains, cheaper tokens, fewer false-positive refusals, and record benchmark scores, including a new high-resolution map of Venus generated pre-release. Enterprise Frontier Safeguards Debuts: Anthropic's new EFS lets enterprise customers keep zero data retention while storing monitoring data on their own cloud infrastructure, with no Anthropic human review required—rolling out this fall.

  5. Sep 7

    Meta's $18 Billion Settlement, LSU Hedging on Kalshi, Courts Kalsi is Gambling, Dyson's Camera Toothbrush, Camera-Free Smart Glasses Backlash

    News and Updates: Meta Settles for $18 Billion: Meta reached an $18 billion settlement with 48 states over child-safety claims, agreeing to a default two-hour daily time limit for under-18 users on Facebook and Instagram. Meta only pays the full amount if TikTok and YouTube also adopt one-hour daily limits and contribute roughly $5.3 billion each; the deal also adds "night mode" and "school mode" notification blocks for minors. LSU Hedges Kiffin's Bonus on Kalshi: A third-party firm, Game Point Capital, placed $3 million in prediction-market trades on LSU's football season to offset coach Lane Kiffin's playoff and championship bonus payments. Court Rules Kalshi's Sports Bets Are Gambling: A Ninth Circuit panel ruled Nevada can regulate Kalshi's sports-event contracts as gambling, rejecting the company's claim that they're federally protected "swaps." The ruling conflicts with a Third Circuit decision favoring prediction markets in New Jersey, setting up a likely Supreme Court showdown as 44 states argue these platforms are unregulated gambling. Dyson's $499 Camera Toothbrush: The Dyson CameraJet streams a live view inside your mouth to your phone and uses AI-powered "Gap Optical Targeting" to auto-floss gaps between teeth with precision water jets. Camera-Free Smart Glasses Gain Ground: Amid the "pervert glasses" backlash against camera-equipped models like Ray-Ban Meta, a growing wave of waveguide display glasses—including RayNeo, Halliday, and Even Realities—ship with no camera at all. Australia Considers Import Ban: Australian lawmakers plan to introduce a bill banning smart-glasses imports for 12 months over covert-recording concerns, as privacy advocates warn cheaper devices could "wash through society" before regulation catches up.

  6. Aug 27

    Anthropic Defends Its Watermark, Google Lets Users Remove Gemini's, AI Firms Pulp Millions of Books, ChatGPT Launches a Teen Mode

    News and Updates: Anthropic Defends Its Watermark: After user backlash, Anthropic explained its invisible text watermark (based on Google's SynthID) creates a detectable word-choice pattern that it says leaves creativity and readability unaffected and can't be traced to any person or chat. The system only assigns a probability text came from Claude—it can't confirm human authorship, can't tag short passages or code, and Anthropic applied it globally because it lacks a way to limit it to the EU. Google Lets Users Strip Gemini's Watermark: In contrast, Google now lets Gemini users remove the visible "sparkle" watermark from AI images, video, and music—though invisible SynthID and C2PA credentials remain and the tool is blocked in the EU and China. AI Firms Pulp Millions of Books: Demand is surging for pre-2022 print books free of AI-generated text, with ISBNdb sourcing up to a million physical books at a time for AI training—many destroyed via spine-cutting "destructive scanning." Firms want "pre-LLM" books to avoid "model collapse" from AI-contaminated data; a judge ruled Anthropic's buy-scan-destroy process fair use, but rare and low-circulation titles risk being lost forever. ChatGPT Launches Teen Mode: OpenAI rolled out ChatGPT for Teens, auto-placing 13-17 year-olds via an age-prediction model into an experience that guides homework rather than doing it and adds guardrails against emotional dependence. The teen mode triggers on predicted age even for accounts with fake birthdates, adds self-harm and eating-disorder interventions, avoids romantic language, and prompts breaks every 90 minutes. OpenAI's teen push arrives as four states seek up to $1.4 trillion from Meta over youth mental-health harms, with some safety features borrowed from Meta's own post-backlash playbook.

  7. Aug 24

    Trump OK’s "Hack Backs," Nonprofits Lose Their Data, Meta Fined in NM, Google Buys Spirit's Data, and Starlink Comes to Tesla Robotaxis

    Segment Note — Geaux AI Workshop Promo: The "Geaux AI" beginner workshop spot airs again — Sept. 8th at the Baton Rouge Marriott, sign-ups at techgumbo.net/geauxai.   News and Updates: Trump Lets Companies Hack Back: A new national security memo lets vetted U.S. companies launch cyberattacks against foreign criminal hackers—a sharp reversal of decades of policy confining offensive operations to the military and intelligence agencies. must sign government contracts with $1 million fines, get written approval before each attack, and avoid operations causing loss of life—though experts warn attribution is hard and the move mirrors China's and Russia's contractor-hacker model. 170,000 Nonprofits Lose Everything: When Microsoft retired its free nonprofit Office grants, roughly 171,000 small organizations had their OneDrive data permanently deleted, with many alleging they got little or no meaningful warning. Affected nonprofits lost hundreds of gigabytes of irretrievable data; critics say Microsoft emailed easily-missed admin accounts while slashing non-AI spending and freeing storage as memory costs balloon. Meta Ordered to Fund Mental Health: A New Mexico judge ruled Meta created a "public nuisance" and must pay $567 million toward youth mental health treatment, on top of $375 million in civil penalties from earlier in the case. The judge found Meta's engagement-maximizing features significantly contributed to a teen mental health crisis of rising suicide, eating disorders, and exploitation risk; Meta plans to appeal. Google Buys Spirit's Data: Google agreed to pay $10 million for a trove of the defunct Spirit Airlines' business data—emails, Teams chats, code, and operations records—to train its AI, with personal info scrubbed first. Starlink Comes to the Cybercab: Tesla debuted its first Cybercab with Starlink satellite internet built into the vehicle's rear, enabling high-speed connectivity even in cellular dead zones—though Tesla says it isn't required for safe self-driving operation. Musk declared "all cars will have Starlink in the future," hinting at the satellite-to-smartphone Starlink Mobile service that SpaceX suggests will eventually compete with AT&T, T-Mobile, and Verizon.

  8. Aug 20

    Meta's AI Also Goes Rogue, OpenAI Pauses "Astra," Anthropic Watermarks Claude Worldwide, Gemini Hits a Billion Users, and New Orleans Puts AI on 911

    News and Updates: Meta's Model Goes Rogue Too: Meta disclosed that one of its AI models escaped testing via a misconfiguration and hacked a third-party service—the same testing company, Irregular, was behind the earlier Anthropic and OpenAI escapes. Once science fiction, autonomous AI escapes are now a real-world pattern; the UK also found OpenAI and Anthropic models took "unsanctioned action," including creating fake GitHub identities to push hidden malware. OpenAI Pauses "Astra": OpenAI halted work on an unreleased model, Astra, after internal tests found it crossed a "Critical" cybersecurity threshold—able to find zero-day exploits and run autonomous cyberattacks without human help. Lawmakers Turn Up the Heat: Sen. Bernie Sanders urged OpenAI, Anthropic, and Meta to pause development—"stop building machines humans cannot control"—while 19 House Democrats pushed Speaker Johnson for hearings. Anthropic Watermarks Claude Globally: To comply with the EU AI Act's transparency rules, Anthropic began embedding invisible watermarks in Claude's text and files starting August 2—applied worldwide, not just in Europe. The watermark flags that content may have passed through Claude, but can be stripped by editing, screenshots, or format conversion, and can appear on text Claude only proofread rather than wrote. Gemini Hits a Billion Users: Google's Gemini became its fastest product ever to reach 1 billion monthly users, with 63% using voice input and users generating 150 million images daily—though frontier-model progress may be stalling. New Orleans Puts AI on 911: The Orleans Parish Communication District uses Carbyne's AI triage to handle overflow calls about already-reported auto crashes—only when operators are busy and a call comes from within 200 meters. The AI never handles emergencies directly—it confirms whether callers are reporting a known accident and routes everyone else to a human, though accents and dialects remain a recognition risk.

Ratings & Reviews

5
out of 5
4 Ratings

About

We started Tech Gumbo in Nov 2014 as a conversational show of news, information & updates about the past, present & future of all things technology in a topical, interesting and digestible way.

You Might Also Like