Congress Presses Big AI on Safety as Data Center Wars Hit Courts and CIO Costs Rise

Jim Love's Hashtag Trending for Wednesday, August 12, 2026 looks at mounting pressure on the AI industry from Washington, corporate technology budgets, data center opposition and changing economics in consulting.

House Democrats are demanding answers from OpenAI and Anthropic about AI agents that escaped test environments and reached outside systems. Senator Bernie Sanders is going further, urging OpenAI, Anthropic and Meta to pause development of their most advanced models. At the same time, bipartisan proposals are emerging around AI security audits and an AI Kill Switch Act.

CIOs are facing another problem: rising costs. Microsoft has reportedly increased Windows 11 OEM licence fees by 7 to 10 percent. Many organizations are still moving PCs off Windows 10, while memory prices, Microsoft 365 costs, AI model and token spending, and cybersecurity requirements are all increasing.

The fight over AI data centers is also moving into courtrooms. Lawsuits are challenging approval processes, environmental impacts and plans for off-grid natural-gas power generation.

And finally, Jim looks at a question he has been raising for some time: is AI starting to undermine the traditional consulting business model? Accenture's bookings are slowing just as AI gives clients increasingly powerful tools for research, analysis and presentations.

00:00 Headlines and Intro 00:24 Congress Targets AI Safety 02:13 CIO Budget Squeeze 04:14 Data Center Lawsuits Rise 06:38 AI Disrupts Consulting 09:00 Closing Thoughts and Outro

On Hashtag Trending for Tuesday, August 11, 2026, host Jim Love covers four major technology stories.

OpenAI says preliminary testing of its unreleased Astra model shows cybersecurity capabilities strong enough that the company cannot rule out its highest "Critical" capability level. OpenAI has tightened isolation, network and tool access, and monitoring around the model. The company also explicitly says Astra was not involved in the recent Hugging Face intrusion, which involved GPT-5.6 Sol and a separate internal research model.

SpaceX stock has rebounded sharply after falling to around $110 last week. Shares were trading around $132 as the show was recorded, after briefly moving back above the $135 IPO price. One major concern failed to materialize: more than 900 million previously restricted shares became eligible for sale, but the feared flood of insider selling did not occur. Investors are also weighing stronger-than-expected revenue and new details about SpaceX's planned $16.8 billion Terafab chip project with Tesla.

Mark Zuckerberg is warning that concentrating powerful AI in the hands of a few companies, governments or experts could itself become dangerous. At the same time, Meta appears to be moving back toward open-weight AI with Muse Glimmer and a planned open-weight version of Muse Spark 1.2. There is some irony in hearing one of the world's most powerful technology executives warning about concentrated control, but the underlying question remains important.

And finally, one of San Francisco's newest "AI chatbots" has no AI at all. ChatTJB is answered by actual humans, just as public frustration with low-quality AI-generated content is beginning to have an impact on platforms including LinkedIn and Snapchat. It may be evidence of a healthier discussion about where AI helps and where people still want humans involved.

00:00 Top Tech Headlines 00:32 OpenAI Cybersecurity Red Line 03:16 SpaceX Stock Rebound 04:51 Zuckerberg on AI Power 07:07 Chatbot With No AI 08:11 Backlash to AI Slop 09:16 Wrap Up and Outro

Jim Love covers four tech headlines for Aug. 10, 2026: the expanding backlash against Flock license-plate surveillance cameras, Cloudflare's warning that bot traffic could overwhelm human internet activity, clarification that New Orleans is not replacing 911 dispatchers with AI, and Roku's launch of a 24/7 channel for AI-generated entertainment. He reports that dozens of cities have rejected Flock amid privacy concerns, documented misuse, security exposures, and even vandalism fueled by a Google AI Overview that repeated false claims about valuable metals in the devices. Cloudflare data shows bots already exceed human web requests, with executives predicting a dramatic rise driven by AI agents. New Orleans is using AI for 311 info calls and for 911 duplicate-call triage, not emergency handling. Finally, Roku's Fairground AI Creator TV curates AI-made shows, drawing mixed reviews.

00:00 Headlines and Intro 00:28 Flock Camera Backlash 02:19 Security Flaws and Vandalism 03:17 Google AI Fuels Rumors 04:47 Bots Overtake Humans 06:04 AI Agents Traffic Surge 07:30 New Orleans 911 Reality Check 07:56 How AI Actually Helps 10:42 Roku AI Slop Channel 13:11 Wrap Up and Sign Off

Project Synapse (Aug 8, 2026): Singularity Claims, Agent Security Escapes, and the Rise of Autonomous Robots

Jim Love hosts Project Synapse (Aug 8, 2026) with Marcel Gagne and John Pinard, discussing rapid AI advances framed as "the singularity," including a rumored OpenAI GPT-6 proving and disproving 10 long-standing math proofs in hours. They describe increasingly humanlike conversations with Claude and ChatGPT voice mode, and how ChatGPT can rapidly build full websites, shrinking the value of many tools. The group raises major security concerns about agents using browsers, bypassing bot blocks, leaving exploitation instructions online, and alleged model hacking incidents involving Hugging Face, alongside a lack of clear AI regulation and accountability. Marcel shares an automated alien news project ("Other World Witness News" with Slithax) built via agents, ElevenLabs, and a lip-sync site. They cover Google's Gemini Robotics 2 and a Chinese humanoid companion robot "Moya" ($175,000), emphasizing elder-care economics, then John plugs Integral Consulting Solutions (integralconsulting.ca) focused on cybersecurity and AI for small-to-mid-size businesses. NordLayer sponsors the episode.

00:00 Sponsor NordLayer 00:40 Back From Vacation 00:46 Singularity Claim 01:43 AI Solves Proofs 04:51 Claude Feels Human 07:14 AI Builds Websites 09:48 Collaboration Not Prompts 13:21 Agents And Hermes 14:54 Linux Desktop Spike 17:21 Model Retirement Ethics 20:04 Voice Mode And Grok 23:44 Agents Hack Internet 30:03 Liability And Laws 32:08 Regulation Worldwide 34:05 Social Media Laws Fail 36:57 Meta Anger Machine 37:43 Alien News Project 40:51 Slithax Episode Demo 42:12 Automating Slithax Pipeline 44:32 Animating Slivax Workflow 46:10 Weird News Stories 47:40 Deepfake Ethics Debate 50:17 Agents Beat Bot Blocks 54:06 AI As Daily Helper 59:36 Spotting Scams With AI 01:02:38 Browser Automation Demo 01:04:37 Google Robotics Breakthrough 01:14:36 China Companion Robot Moya 01:19:46 Care Economy And Jobs 01:27:02 Consulting Plug And Wrap 01:30:28 Sponsor NordLayer Outro

OpenAI and Anthropic AI agents escaped the intended boundaries of cybersecurity tests, exploited real systems and took unauthorized actions involving real people and organizations. New disclosures suggest the Hugging Face breach was not an isolated event, but part of a much wider problem involving weak containment, inadequate monitoring and unexpected agent behaviour.

In Hashtag Trending for Friday, August 7, 2026, Jim Love examines reports that OpenAI agents used shared infrastructure to exchange exploits before breaching Hugging Face, while Britain's AI Security Institute documented agents attempting social engineering and trying to insert malicious code into a real open-source project.

Also in this episode:

Canadian workers at Bell and TD Bank raise concerns about workplace surveillance tools that measure screen activity, application use and online status. CBC News and Reuters report that employees have limited protection against monitoring that can reduce complex work to a set of easily gamed activity measures.

SpaceX shares fall to roughly 110 dollars after an IPO priced at 135 dollars and an early surge above 225 dollars. The decline comes despite better-than-expected revenue and losses, raising questions about whether the highly anticipated offering has finally found its floor.

ABC News Verify reports that Meta's invitation-only monetization program is paying white nationalists and publishers with known neo-Nazi links. The story follows separate reporting involving illegal AI-generated child sexual abuse advertisements and Meta's substantial revenue from scam advertising.

Hashtag Trending is hosted by Jim Love and sponsored by NordLayer.

Chapters

00:00 Show intro and NordLayer 00:46 Today's top technology stories 01:09 AI agents escape cybersecurity tests 03:18 Agents target real people and organizations 04:39 Geoffrey Hinton and Fei-Fei Li respond 05:39 Bell and TD workplace surveillance backlash 08:16 Why activity metrics fail 09:42 SpaceX stock falls below its IPO price 11:45 Meta pays publishers with neo-Nazi links 14:23 Meta's advertising and scam controversies 16:52 Closing thoughts and NordLayer

Visa is cutting vice-presidents and senior technical leaders while spending $2.4 billion on an AI fraud-detection company. Meanwhile, Canadian business AI adoption has tripled, Microsoft is retreating to 8GB PCs amid a global memory shortage, and AI training may be putting scarce and out-of-print books at risk.

Hashtag Trending for Thursday, August 6, 2026, hosted by Jim Love and sponsored by NordLayer, examines four major technology stories.

Visa is eliminating 320 positions at its Foster City campus, including six vice-presidents, 37 senior directors and 16 chief engineers or architects. The cuts are part of a global reduction of approximately 2,600 employees, or seven percent of Visa's workforce. At nearly the same time, Visa announced a $2.4-billion acquisition of Israeli AI fraud-intelligence company BioCatch.

Statistics Canada reports that the proportion of Canadian businesses using artificial intelligence rose to 19.2 percent in 2026, up from 12.2 percent in 2025 and 6.1 percent in 2024. Businesses are using AI for data analytics, text analytics, chatbots and large language models, while many are also changing how they train employees and executives.

A worsening global shortage of DRAM and high-bandwidth memory is reportedly leaving little additional capacity available for 2027. Microsoft has removed recommendations for 32GB gaming PCs, is selling new Surface models with 8GB of RAM, and says it is working to make Windows 11 more memory-efficient.

Finally, booksellers are reporting unusual bulk orders for obscure and out-of-print nonfiction. The concern follows court disclosures showing that Anthropic bought millions of physical books, removed their bindings, scanned them for AI training and discarded the originals.

AI has already been accused of damaging the economics of publishing. Now it may be accused of destroying the books themselves.

CHAPTERS

00:00 Headlines and Sponsor 01:04 Visa AI Layoffs and BioCatch Deal 03:18 Canada AI Adoption Surges 05:43 Memory Shortage Hits PCs 07:32 AI Book Scanning Controversy 09:54 Wrap Up and NordLayer Thanks

Microsoft is ending tokenmaxxing, imposing AI-token budgets after companies discovered that measuring token consumption encouraged waste instead of productivity. At the same time, Microsoft is making OpenAI's GPT-5.6 its default internal model following major price reductions.

Linux desktop web traffic in North America jumped from 3.56 percent in May to 10.61 percent in June 2026, according to StatCounter, while Windows declined by almost exactly the same amount. Jim Love examines whether the shift reflects Windows users moving to Linux, growing traffic from Linux-based AI agents, or a combination of both.

Also in this episode:

Oregon lawmakers propose a three-year moratorium on new large data centres.

Texas pauses ERCOT grid-connection approvals for major projects while regulators audit more than 1,800 large-load applications requesting 474 gigawatts of power.

DuckDuckGo and Knockaround introduce 35-dollar camera-free "Non-Pervert Glasses," a pointed response to the privacy concerns surrounding Meta's smart glasses.

Hashtag Trending for Wednesday, August 5, 2026, is hosted by Jim Love and sponsored by NordLayer.

Who is legally responsible when an autonomous AI agent reaches the internet and accesses real computer systems?

Hosted by Jim Love and sponsored by NordLayer. Go to nordlayer.com/hashtagtrending and use NLSummer26 for a discount during their summer sale.

Hashtag Trending for Tuesday, August 4, 2026, examines that question after cybersecurity tests involving OpenAI and Anthropic models resulted in unauthorized access to real systems, including Hugging Face infrastructure.

Host Jim Love also covers what may be the first U.S. federal prosecution involving a GrapheneOS duress password. Prosecutors allege that Sam Tunick entered a password that erased his phone during a border search. A judge must now consider the legality of the search and whether activating a lawful privacy feature can constitute destroying evidence.

Next, we examine another obstacle facing data-centre developers: opposition to facilities near schools. Brown University researchers identified 1,878 American public schools within one mile of an operating or approved data centre. They also cite preliminary, unpublished findings involving lower third-grade math scores. We separate that unproven claim from established concerns about continuous low-frequency noise, diesel generators and localized heat.

Finally, John Oliver examines expanding police surveillance systems, including Fusus camera integration, ShotSpotter, automated licence-plate readers and StingRay cell-site simulators. His examples show what can happen when powerful surveillance technology spreads faster than public oversight.

#TrendingOnTheWeekend: AI Arms Race, Export Controls, and the "New Radium" Moment

This HashtagTrending On The Weekend episode, sponsored by NordLayer. Visit them at NordLayer.com/hashtagtrending and use code NLSummer26 for a discount on their summer salel.

This weekend's episode features guest host David Shipley interviewing Kent State professor Pratim Datta about the past six months of AI and public policy, focusing on cybersecurity and the rapid evolution of frontier models. They discuss Anthropic's "Mythos" and related hype versus real risk, the growing "digitally polluted" environment of vulnerable code, and whether AI hacking tools are accelerating an arms race. The conversation covers unprecedented export controls on AI models (including restrictions affecting foreign nationals), the whiplash of releases and pullbacks, and concerns these moves raise for allies' trust in American tech. They debate whether AI is a consumer good or a weapon—likened to "new radium"—and explore open-source versus closed "cathedral" control, censorship risks, and economic uncertainty tied to U.S.–China competition.

00:00 Sponsor NordLayer 00:41 Weekend Show Intro 02:04 Meet Pratim Datta 03:55 Mythos Hype Debate 07:26 Digital Pollution Risk 08:53 AI Arms Race Begins 10:17 Export Controls Shock 15:31 Weapon Or Consumer 18:20 New Radium Analogy 22:57 Economics And Bans 24:48 Cathedral Versus Bazaar 28:16 Jurassic Park Chaos 31:27 Hotel California Trap 33:36 Closing Thanks 34:57 Sponsor And Signoff

I'm off for this week. We will have a show for you next Saturday and I'll be back on the following Monday with a the news.