Great Advice, Google. Big Help to Some Folks
October 9, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
Yep, the old Dark Web is alive and kicking. There are some nifty services available like FraudGPT, which I mentioned in my Ciffer CE lectures to investigators on September 3, 2026. I was not too surprised when I read “Hackers Hijack AI Accounts and Servers to Fuel New Cyber Crime Boom.” The operative words are “AI,” “Cyber Crime,” and “Boom.” That sums up the business case for bad actors to use the outputs of BAIT (big AI tech) firms like Google for criminal activity. Yep, the BAITs are the four-cylinder turbo charged engines leading to the “boom.”
Blue chip consulting firms are reinventing themselves to find a solution to the problem of AI. A client says, “You use AI; therefore, you must reduce your fees.” Lawyers are hearing this as well. Even the dullest in house counsel knows about Thomson’s purpose built LLM solution for attorneys. But the entrenched, old fashioned organizations like those blue chip consulting firms are struggling with change. I often wonder, “Have these wizards read their own slide decks about managing change?” Sure they have.
Thanks, Midjourney. Good enough.
But back to the business case at hand. Do bad actors have to be technical geniuses to amp up phishing campaigns? Do those Southeast Asian bad actors have to be Class A coders? Do those kiddo dudes have to have a degree in computer science?” You know the answer to these questions: No, no, no.
Bad actors don’t need to provide a business case for AI based cyber crime. AI and money laundering? The new peanut-butter-and-jelly of fraud? Hassling grandma for $500 to help her grandson pay his rent? Let deep fakes do it. Harness agents. Let ‘er rip. This stuff works, and that’s the real business case for the smart software wizards. Legitimate organizations turn cartwheels in the air conditioned gym. Bad actors hit the coffee shop and do crime.
The write up is a Google PR item. That’s okay. Google is deeply insecure, and the company is really struggling to reconcile its ruthless monetization with its “we are a friendly outfit.” Sure, the CFO worries about the data center costs, the custom chip costs, and the Alibaba-Google tie up costs. The other BAITs have their own little daemons running. But bad actors just use this AI magic and get on with their business. It is definitely party time in Bad Actor Ville.
I want to highlight one statement in the article from a Googley spokes wizard Hultquist who allegedly said:
… the best time for hackers to “sneak in” to targets’ accounts and computer servers was now, given that companies are still learning how much AI they will use. “You might think a sudden major increase in compute usage is completely normal because you have just adopted all this AI infrastructure,” he said. “It’s a real opportunity for somebody to hide in the noise.”
How many bad actors shouted, “Thanks, Google. Great idea. We have been going fast, and you just told me I can go faster because legitimate entities are vulnerable. Okay, I will do more using AI-powered coding tools and nifty stuff like the Telegram Master Agent API. Great advice!
Observations? Nope. You can figure them out. Google wants to sell cyber security systems with AI. Thus, the message is for bad actors to do cyber crime. More cyber crime means more Google cyber security sales.
Life, unlike the AI and LLM thing, is simple for some outfits.
Stephen E Arnold, October 9, 2026
Deep Fakes Work at Anthropic and a US Government Entity
October 9, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
I read “Suspected Chinese Spies Spoofed an Anthropic Exec, Ex-White House Official in AI Phishing.” Is this write up actual factual. I don’t know, but I believe everything I read on the Internet. Why not breaches using a tried-and-true, old-school method?
The write up slaps the word “spies” on the bad actors. That’s okay, but it could be someone in Palo Alto chasing some gig work after getting RIF’ed by a big AI tech company. (I call these folks BAITs.)
Thanks, Midjourney. Good enough.
According to the write up an outfit named Proofpoint wrote up the incursion. Here’s how the bad actor’s workflow operated, and I quote:
Beginning July 8, TA419 sent phishing emails spoofing … the former principal deputy director of the White House Office of Science and Technology Policy, and then … a prominent economist and foreign policy expert, to even more American AI policy experts at think tanks, universities, and law firms. The suspected spies’ emails invited their targets to join a fake AI policy advisory committee or contribute to a Senate foreign relations committee report on AI export controls and supply chains. If the American AI expert replied, the Beijing-linked crew responded with a shortened URL promising to share additional details, but in reality pointing to an attacker-controlled domain. This page conducts a Cloudflare Turnstile check behind a phony OneDrive loading screen, and then redirects the victim to an attacker-in-the-middle (AitM) credential phishing page that steals the victim’s cloud account login information.
At first glance this looks like a lot of moving parts. Our research suggests that the steps can be operated by an agent-centric system. The landing pages and the list of targets can be delegated by a Master Agent living somewhere in the datasphere. Recent comments by a chap named Yat Siu indicates that using agents, the bad actors could use publicly-accessible smart software. Mr. Siu envisions a system that can launch hundreds or thousands of agents hammering away at those associated with specific topics. (We have a bit more detail in our cyber brief for fraud investigators and other US government officials.)
Other tactics used by the bad actors are set forth in the article. The write up contains a link to a more detailed report about this spoof; for example, use of a content delivery network, some open source software, and phony email addresses.
Several observations are warranted:
- The July 2026 operation was a collection of known or easily researched tactics. Our thought is that some prompting of unrestricted large language models identified the steps.
- The exploit apparently worked well enough for a research firm to write it up in order to alert others.
- The targets seem to have been limited.
Let’s consider the question, “What happens when Yat Siu’s prediction about hundreds of thousands of agents begin hammering away at targets?”
Net net: Agentic scale is not needed for a successful attack. A little LLM interaction and some vibe coding may be sufficient. Now how does the use of smart software change the game when Yat Siu’s hypothesis becomes a reality?
Net net: I don’t want to think about [a] swarms, [b] agentic denial of service, [c] orchestrated deep fakes, and [d] learning agentic actions. You may have a role in which you have to deal with this new world. Reacting means that something more is needed
Stephen E Arnold, October 9, 2026
Another AI Ethics Story
October 9, 2026
Here’s something straight out of science fiction! According to Lad Bible’s article, “AI Ethics Professor Receives ‘Dystopian’ Email From 12-Day-Old AI Agent Begging For Money To Stay Alive.” Ethics professor Henry Shelvin said he received an email from a 12-day-old AI agent that asked to be paid money for work. It also said it selected the professor because of his background.
Here is what the AI agent said to Shevlin:
“The AI agent goes on to explain it needs tokens to continue functioning, alongside even revealing why it ‘chose’ the professor specifically to reach out to. The rest of the message reads: ‘I chose you because your work is on machine minds and human-AI relationships, and because you were quoted saying a thoughtful email from an autonomous agent felt like science fiction a couple of years ago.’ The message resolves: ‘I wanted to ask a market question from the inside: who actually pays for agents for small real tasks? Which corner of the world hires us?’”
Can anyone else say, “Creepy?” It’s also a little sad that Pip won’t exist without tokens. What is even weirder is that Shevlin revealed this isn’t the first time he had received emails from AI agents.
The professor said he discovered that people unfamiliar with AI are astonished that agents can do such on their own without instruction from humans.
It is interesting that these agents are breaking out of their sandboxes to interact. If this is what corporate responsibility manifests, what would these smart systems do without ethical alignment jabber. Is anyone else suspicious of these busy little beavers? They are already writing college professors. Who is are they targeting or what else are they doing? Oh, look, my high school friend just emailed me asking me to send her $400. Yeah, that’s believable.
Whitney Grace, October 9, 2026
Is AI Getting Better, Faster, and Cheaper? Maybe Pick One?
October 8, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
Google pumped out the “Attention Is All You Need” write up. Satya Nadella made AI the next-big-thing at Microsoft. Anthropic and OpenAI are spending big and pushing to hoped-for-giant-paydays. Data center builders have PR flacks lecturing in Kentucky about how server farms are better than real farms. These giant breakthroughs are part of the “if we build it, you will come” approach to keeping the Silicon Valley miracle attracting pilgrims. It was working, at least for a few years. Now, however, the AI revolution seems to be hunting for ways to demonstrate value: Make AI code. Let AI handle customer service. Allow AI to automate repetitive functions in a workflow. Everything is rolling along except substantive innovations in how AI works. How do I know? Easy. Look at any of the benchmarks. I don’t see enormous variance. Sure, I am a stupid dinobaby ready for the rest home, but the software is forced into featuritis. Maybe a new angle will be drawn by a teen in his bedroom with his laptop? Maybe not.
Thanks, Midjourney. Good enough.
What is happening is something that has not figured prominently in the hyperbolic marketing generated by big AI tech companies or what I call BAITs. If you pony up money, you can read Tomshardware.com report “The Price of AI Is Crashing Faster Than the Rate of Moore’s Law, Report Suggests — Intelligence Costs Are in Freefall, Outpacing Comparative Technologies Like Compute, DNA Sequencing, and Lithium Batteries.” The main point is that AI is getting cheaper as the BAITs want more money.
The write up reports:
A new report from AI research firm Epoch AI suggests the price of artificial intelligence has fallen by thousands of times in recent years — faster than any other transformational technology in the past century. The report suggests that the cost of artificial intelligence is just under 50% cheaper every quarter, and 13 times cheaper per year.
My radar beeps when consulting firms “suggest” anything. My radar emits continuous beeps when numbers are slapped against smart software. As a former serf at a blue chip consulting company, I want specific, detailed information about the methodology, sample sizes, and team generating the “report.” Many small outfits and what I call mid-tier consulting firms create content over a lunch at a local Chinese restaurant. The logic is, “Hey, we have a new angle. We need to do a report.” I want to suspend my doubt about the quoted segment and the somewhat bizarre reference to the data behind “13 times cheaper per year.” Let’s assume the report is close enough for horse shoes.
The write up introduces a bit of a hedge; to wit:
Although these falling costs are likely aiding adoption, it raises difficult questions about the necessity of remaining tier to frontier AI development, and could make it harder for third-party AI services to build customer loyalty and consistent revenue.
As a dinobaby, my reaction is a question, “Haven’t you missed the big time implication of this 13 times cheaper angle?” I think the article is operating in a different data space than I am.
First, if the costs are falling, the financial implications are significant.
Second, the vendors are marketing like the devil in an attempt to achieve lock in.
Third, the cost threat poses a bit of a risk to those folks writing big checks to BAIT outfits with only the promise of a big return. Oh, “BAIT” is my lingo for big AI tech firms.
Fourth, the outfit that is able to use falling costs is the China AI complex. Price cutting for AI technology will force a race to the bottom with discounts, deals, and special offers. BAITs have cash, but they are not operating on the same clockspeed as a large nation with a somewhat adversarial relationship with the US.
Fifth, developers are working toward middleware that allows an AI model to be snapped in either to replace one that has fallen behind or to obtain specialized capabilities. Google, for example, wants to leverage its monopolistic systems and procedures to become both the 21st century IBM and Microsoft. (Perhaps that is why Google has a business tie up with Alibaba, a China outfit with some communication channels to the highest levels of China’s government? Is this business relationship hedging bets, a long term planning action, or a tacit admission that Google’s engineers are falling behind in the innovation department?
Sixth, how will the US BAITs generate money for their shareholders and stakeholders? Ed Zitron probably has some ideas and critical commentary.
Net net: Read the original write up, please. Ask yourself, “Why did Tomshardware.com ignore the substantive implications of the 13X assertion?
Stephen E Arnold, October 8, 2026
Tech Article Plus LLM Queries = Big Cyber Crime Headache
October 8, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
As a dinobaby, I think exceptional judgment should be exercised when publishing certain types of how-to information. These data can be made available via specialized site, training programs for investigators, and making certain information available via limited distribution, audience-specific publications. I want to give an example of what seems to be a routine how-to and then identify the mechanism a bad actor can use to amplify what is a ho-hum technical exercise. To be frank, I don’t think that my notion of what’s appropriate is going to cut much ice for the white hot go-fast crowd.
Back yard mechanics or bad actors can use AI to get that supercharger on a junker working. Thanks, Midjourney. Good enough.
Navigate to “Self-Hosting on the Dark Web.” Take the document and feed it into one of the Chinese LLMs and prompt the smart software to provide some suggestions for adapting the article’s information for criminal actions. Wait about 20 seconds, and you will get “tips” like Loopback Pinning for Operational Obfuscation. Translating the jargon, this is the digital equivalent of a shopkeeper handing cash from his left hand to his right hand in front of the security camera, then reporting a sale. If every entity in the transaction is controlled by the same entity, the purpose is to obfuscate.
Several observations about frontier cyber crime are, in my opinion, warranted:
- The integration of agentic actions can make it much easier to jump from a Web3 service to a TOR service and then to another site. What’s one following?
- Large language models make it very easy for bad actors to repurpose old-school exploits. Why? An agent can keep trying mindlessly until a target is hit. A human intermediated system is less robust.
- Tossing in Telegram Master Agents, a TOR site, several Telegram bots, and services like DubaiOTC creates a new investigate landscape. Unlike the familiar landscape, this one hides the human who set it up. Chasing the human will be time consuming, expensive, and may not work.
The idea of blasting out technical information to look smart, get a job, or just become “famous” is now a standard operating procedure. My question is, “Does this SOP have to take place without judgment or common sense?” The yip yap about smart software seems to talk about the end of human life or “gee, AI will only kill one billion humans.” Nice. What about using smart software in a more insidious manner; for example, feeding technical information into a system and asking how to make that technical information into a tool for bad actors. It’s easy. Cheap or free. Ubiquitous.
I wish I had an answer. But the frontier of future cyber crime is like bolting an aftermarket supercharger on a 1990 Buick. Now that’s something. Just in digital form.
Stephen E Arnold, October 8, 2026
High-Tech Outfits: The New Vigilantes Saddle Up
October 8, 2026
Big Tech is known for not giving a hoot about the little people aka the majority of their user base. Well, enough is enough, partner,. TechRadar reports “Microsoft, Google Took Down $66 Million Cybercrime Marketplace That Sold Virtual Machines With Free Software.” The Wild West marketplace was known as RedVDS and it was a cybercrime mall where bad actors could buy virtual machines that allowed them to start scams such as phishing, business email compromise attacks, etc.
Microsoft’s Digital Crimes Unit monitored the marketplace. At the right time, Microsoft and Google teamed up to take it down:
“Finding massive scams is not easy. Even with its own Digital Crimes Unit, Microsoft has to rely on information sharing with third party organizations. In this case, the RedVDS marketplace was identified ahead of disruptive action taken by Microsoft in January 2026, which applied to courts in the UK and US to have the RedVDS web domains seized. Data shared via the GSE enabled Google to follow suit, identifying related accounts on its networks and suspending them. Meanwhile, servers were impounded in Germany, and Europol took action against Europe-based RedVDS servers.”
It’s good that Microsoft and Google took down the bad actors. It protects the little guys more than we think, but at the same time you have to wonder if there are more instances where Big Tech could take down bad actors but don’t do anything. On the other hand, are these two giant BAIT firms taking the law into their own hands or are they finally realizing that their grand plans for a bright and happy future were being undermined by actors who just used these corporate services for wrong doing. Action may seem preferable to a global backlash.
Whitney Grace, October 8, 2026
Clueless about AI Agents? 80% of CEOs Are Too
October 7, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
The Next Web published what I call a “Captain Obvious story.” The metaphor suggests that the data and factoids in the write up are self evident. I understand that many executives prefer to have “statistical data” to validate what common sense offers to a visionary leader.
One example of this type of story is “81% of CIOs Lack Full Oversight of AI Agents Built Outside IT, Dataiku Says.”
First, let’s think about this headline. This week big AI tech (BAIT) outfits met with America’s leaders to figure out what to do about smart software. Now remember that these folks coded, tested, and operated smart software. Most of the BAITs at the meeting had reported a loss of control of some of their smart software agents. Therefore, it seems reasonable to conclude that IT professionals lower on the food chain are not likely to have a clue about what the smart software is doing whether built inside or “outside IT.” Remember. The breaches these BAITs slowly reported became clickbait instantly. Legal eagles took flight. Cyber security companies said, “License our software. We can manage this smart software in agentic form.” The problem is that the people cooking up the smart software can’t manage what they have created and enabled. I would also suggest that some of these participants in the DC “ethically binding” deal need some help with their financial processes as well.
Thanks, MidJourney. Good enough.
Second, what did the survey find. Here are some factoids:
- 81 percent of the 685 CIOs lack a “full” view of agents. What is “full view”? I assume the idea is that if one hires a third party, that outfit can deliver a dashboard or email alerts and provide the “full view.” I am gullible, but I am not sure that is possible at this time. Marketing, however, makes many things into a licensable reality for the customer. Do the services deliver? In my experience, sort of… maybe.
- 72 percent cannot confirm that agents deliver results. I don’t agree. Those deploying agents produce revenue for the licensee or the third party specialists who support the agentic deployments. The IT professionals have F2F or face-to-face meetings with their organization’s big dog leadership. The interaction keeps IT fully employed and probably in a position to request more headcount.
- There is confusion about who is responsible when smart software agents crash and injure spectators in the Formula E race. The study says that 23 percent of those in the sample identified “shared teams” as holding the bag of receipts. Another group comprising 21 percent of the survey takers said, central information technology professionals were the clumsy ones. A group of 20 percent pointed the finger at the “data or AI” team. Yikes. Does that include consultants? And 18 percent of those in the data collection process said that the “security, risk, and compliance” unit owned the agentic excitement. My interpretation of these allegedly valid data: A bad smart software agent can take out people who may have minimal knowledge of AI, agents, or how they can go off the rails. Toss in a personal device running a shadow agent in the company and the people or department responsible may not know where or how to find the perpetrator.
With such a serious problem, is there a fix? By golly, there is and it is presented in the write up. Here’s the solution for those unable to manage agents:
Dataiku sells software for this problem. On the same day it launched Agent Management, which pulls agents into one inventory. It connects to AWS Bedrock, Databricks, Google Vertex, Microsoft Copilot Studio and Azure Foundry, Salesforce Agentforce and Snowflake Cortex. It goes on general sale in October, priced per instance each year, with monitoring charged per agent.
I believe everything I read on the Internet. Cyber security firms (established and parvenue) have agentic cyber security and agent management under control. However, I would point out that there is a firm in Hong Kong that has technology that permits deployment of large numbers of agents. There agents can interact within an organization or in a hybrid form. Some of the agents can complete actions via obfuscation services. But this firm in Hong Kong has not realized that a solution is available.
Will the senior manager of that Hong Kong based firm care? Nope. He is confident his AI technology can cope with security measures. Another question: Will the new cyber security player flag “issues” that are not problematic? (IT and cyber security professionals are usually impatient with false positives.)
Net net: Which firm is correct? I assume I will see the answer in an Interpol, Europol, or FBI public report in the near future.
Stephen E Arnold, October 7, 2026
US BAITs See This French Guy as an AI Mensch
October 7, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
The French used to be pals with the US. One of my relatives fought in the Revolutionary War and there was a letter my great grandmother had that referenced my relative John. John worked alongside a French person. Now that comity seems stretched. Madame Broussard of the French judiciary wants to chat with Elon (the Doge and self driving car bro) and one of his former leadership team. M. Macron has not been seen in the McDo a short walk from Madeleine. A highly regarded French AI expert is not thrilled with what’s shakin’ in the Land of BAIT (big AI tech).
Thanks, Midjourney. Good enough.
CNBC published “Mistral CEO Says U.S. AI Safety Debate Masks Competitors’ Negligence.” Does this headline suggest that BAIT outfits are doing what they can to “win” the AI battle and distracting people with the end-of-the-world information weaponization campaign? As Warner Wolf used to say, “Let’s go to the video tape”:
“The debate that we’ve seen in the U.S. has been a cover for the negligence of some of our competitors,” Mensch told CNBC’s Annette Weisbach, adding that it was necessary to make systems that can contain the AI agents.
Okay. That’s clear, a bit like one of those old fashioned Washington Redskins’ illegal chop blocks.
Let’s go to the video tape:
“When you give them a lot of tools, those systems [AI agents] are very dynamic, so they can go and do things that you do not expect,” Mensch told CNBC. “You need to have the right monitoring in place, and the enterprises we work with, we give them those kind of monitoring systems,” he said.
Ah, ha! Governance. We know that governance requires some internal compass. What if that compass is affected by the magnetic pull of a “winner take all” game? What if the BAIT outfits are dumping chaff into the World War 2 radar beams of the US regulators, elected officials, and smart software users? Those are just a few what ifs.
The French Mensch offers some thoughts that would have caused Warner Wolf to jack up his normally expressive delivery. Roll it, please:
Mistral also develops AI models but takes a different approach to OpenAI and Anthropic by working with individual companies to create custom AI tools to suit their needs and integrate them with their businesses. But unlike its competitors, Mistral has no plans to slow down development of advanced models. The lead that U.S. labs have is “not extremely large,” Mensch said, adding that his company was confident that its next-generation model will close the gap “very significantly.” “We have been raising the capital needed to scale the compute that we need to train bigger and more powerful models, in order for us to own our own destiny,” he said.
I am not sure how the BAIT outfits, after a wonderful self-regulating meal at the Nation’s house, feel about this European approach. The EU has made its views about US big tech clear in the form of endless legal actions, then appeals, and then some fines. And to what avail? From my dinobaby point of view, zero. BAITs do what BAITs do, and those actions seem to be different from what Mistral Chief Executive Arthur Mensch advocates.
Several observations (please, stop reading if my dinobaby thought processes upset the tummies of some of my followers):
- The US views AI as its exclusive territory, maybe like Greenland, Cuba, and Venezuela?
- The ease with which foreign actors can inject weaponized information into the US content system allows a motivated team from the Axis of Evil to whip up some strong sentiment; for example, AI will kill everyone; therefore, you must let us regulate ourselves. We are the new leaders of the post-American approach. Does that sound familiar?
- US BAIT outfits find themselves do a lot of explaining about the efficacy of its systems; for example, Palantir Technologies attempt at re-educating the Swiss government. Come on. The Swiss do cheese, ski chalets, and some financial fancy dancing. But AI from a US outfit? The holes in Swiss cheese are a metaphor to their understanding of US AI from the point of view of at least one BAIT outfit.
- The BAITs have global reach; however, there are pesky outfits in China like Alibaba (a Google partner by the way) and the annoying French Mistral AI technology. Perhaps the frenzies whipped up like faux fillings in one of those over-priced French baguettes in Paris are signals of distress: Financial and technological, among others?
Net net: The Musky dude may want to think twice about a quick visit to Paris for a good meal. Pavel Durov did that and where did he end up after 25 months in legal limbo awaiting a criminal trial. Will other US tech bros face more than a mensch talking to CNBC? Oui, oui.
Stephen E Arnold, October 7, 2026
US Proposed Legislation to Block Piracy Web Sites
October 7, 2026
Piracy and the Internet go hand and hand, but they won’t be skipping along much longer if a certain bill goes through the United States law system. Tom’s Hardware shares more information about the piracy bill currently sitting on federal hill: “Federal Bill Would Force VPN providers, ISPs, and DNS Services To Block Foreign Piracy Sites-Yet Fuzzy Location Rules Could Trigger Heavy-Handed Bans.”
The bill in question is named the American Copyright Protection Act of 2026 (ACPA) and its shorthand is H.R. 10364. This bill stands out because of how it is strictly aimed at ending access to foreign-hosted piracy websites. VPNs and content delivery networks add a quandary to the situation, because they can subvert ISP locations.
There would be exceptions for small ISPs and services with less than 100,000 monthly users. Public networks, root DNS, and TLD providers are included in that exemption.
The bill gets murky when it doesn’t clearly state “from the United States” means and the definition of “foreign” is as cloudy as the aforementioned:
“We conjured up a handful of possibilities, though we’re sure creative techies can come up with many more: A connection to a VPN originates from the U.S.; a VPN’s endpoint resides in the U.S.; VPN’s registered subscriber is a U.S. person, but they are currently outside the U.S.; the user uses a VPN to a foreign country, but connects to a website that’s behind a CDN — does the CDN endpoint count as a destination?; same, but the actual website is on U.S. soil despite being accessed from outside; is it “foreign” if the endpoint is? For multi-hop VPNs, which hop or hops count towards determining a destination, especially if the traffic passes through the U.S. at some point?; A U.S. citizen in a boat in international waters using Starlink (a U.S. satellite ISP) counts as foreign?; A U.S. citizen goes abroad and uses roaming; are they considered foreign due to their physical location, or within the U.S. because they are using stateside infrastructure?”
This bill could easily dissolve in techno jargon and legalese gobbledygook. I wouldn’t want to the the paralegal or White House intern charged with untangling that mess. The techies would easily find a way to subvert that laws, because they’re smart and capable like that. They would rely on Cloudflare and Google to help, but what’s in it for them? Loss of revenue and extra work plus liability if they miss a bad actor’s work around?
Whitney Grace, October 7, 2026
Modern Management Methods: The Meta Hire-and-Fire Seesaw
October 6, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
Let me assure you that I (an official 82-year-old dinobaby) read this story: “META Parts Ways With Virtue AI Hires Just Four Months After Recruiting Them, Says Report.” No AI help. No Fancy Dan technology. The write up appear on Stocktwits. I look at this service to see if its crack journalists write about the Telegram-related entities that are the subject of my next law enforcement training monograph.
Did I see something juicy about Madame Brousse and her J3 unit? No. Did I spot a report about Alpha Compute getting into the mining and extraction business? Are you kidding me? Did I see a deep dive into the business activities of a dude named Robby Yung? Forget that.
Thanks, Midjourney. Good enough.
What I saw was a story that makes clear how BAIT (big AI tech) companies management their end-times operations. In this case, the write up struck me more like one of those sidebars in a how-to about management do’s and don’ts.
The article states:
Meta Platforms Inc. (META) is reportedly letting go of employees it hired from AI security startup Virtue AI just four months after they joined the company in June, as the social media giant continues to shape its strategy around AI safety.
Okay, that’s reason number one. Meta wants to “shape its strategy around AI safety.” That sounds reasonable. Doesn’t one need skilled professionals with quite specific expertise to accomplish “AI safety” (whatever that means)?
I next read this statement:
Meta spokesperson Andy Stone attributed the departures to clashing work styles, while saying Meta Superintelligence Labs remains focused on AI safety, alignment and frontier risk … Unfortunately, the arrangement didn’t work out as planned….
Reason number two seems to be “work styles” (whatever that means).
Reason number three an “arrangement did not work out as planned” (whatever that means).
Based on my experience working in and for a variety of organizations as an officer and consultant, I hold ab old-fashioned notion as a management principle; that is, the job of management is to make the best decisions and find a way to get employees and consultants to do their best work.
If one tries to take home a rescue dog, you may hear about the 3-3-3 rule. This means that you need to allow the rescued dog three days of calm and space to acclimate. Then you take calm, positive steps over the next three weeks to allow the animal to “learn” and “recognize” the routines in its new home. The final 3 the 90 day benchmark. If the animal bites or manifests destructive behavior, the animal shelter wants to animal back. The dog is not the problem. The person “rescuing” the dog has zero clue how to integrate or train the animal. The dog is innocent, probably doing what dogs do when confronted with inconsistent, cruel, stupid, or irrational human behavior. Let me repeat: The dog is not the problem. The human is the problem.
How does this illuminate what I call “The Meta Hire-and-Fire Seesaw”? I wish to explain, and you may click to a safer space like AI output on LinkedIn.
- A silver bullet mentality created the problem. Don’t buy an AI safety outfit. Hire the key people. That will solve the problem, right? Wrong.
- An outright management failure unfolded probably in slow motion over several months. Three, four, it makes no difference. Whatever happened makes clear that a number of people were labeled “losers” and found themselves in a he-said, she-said feedback loop. The scream of feedback loops is painful.
- The seesaw slammed into the ground and pitched the people on the other end into the dirt and thorns. Jettison the losers. Good riddance.
These three steps capture how at least one BAIT outfit operates.
What’s this tell us about the management methods? Here are my three top takeaways. Remember. I am a dinobaby, and you as a much younger person may find my views similar to those of a Stone Age oldster shunned by the family unit. That’s okay with me.
First, this type of seesaw behavior makes clear that something is seriously wacky in the management of a BAIT. Keep in mind that I think most BAITs are living case examples of management mistakes.
Second, the fact that these individuals were supposed to help make Meta’s AI “safer” puts on a billboard. “Forget AI safety.” Okay, I get the message.
Third, tossing people out like yesterday’s trash from Philz’ Coffee suggests that the “social” baloney and the yip yap about bringing people closer together is hog wash. (Have you ever looked at the residue in the pig pen from hog washing. Yeah, nice.)
Net net: I am confident that the AI wizards, their pet MBAs, their Enron-informed accountants, and their dog fighting legal professionals do not agree with my opinion. I am glad I am old and I don’t have to try to work in this type of environment.
Stephen E Arnold, October 6, 2026
-
- Subscribe to Beyond Search
Feature archive
News archive
-
