#chat-gpt — Public Fediverse posts

Trump’s AI Adviser?

David Sacks starts with an opinion, I agree with: that it is stupid to think AI will end humanity.
‘At least that’s how I remember it…’

https://youtu.be/8ox3tcrUpAc

Prompt Bonsai 27B:
@Web Fact-Checking via Google Search
@video-transcript-downloader
Video Link
Provide your opinion as an Advanced AI Scientist for a Futurist.
1. Review the video in under 500 words and recap key points.
2. Research reports on AI restrictions.
3. Explain how and why AI will advance human lives.
4. In the future, how should AI be moderated for public use?

Bonsai 27B Said:
Advanced AI Scientist for a Futurist: Analysis & Synthesis

1. Video Recap: Key Points

Video: “Extended interview: Trump AI adviser David Sacks on tech leaders’ AI fears” (September 2026)

David Sacks, serving as a key technology advisor to the Trump administration, addresses concerns from the AI/tech community regarding artificial intelligence’s trajectory. His core argument centers on pragmatic optimism about AI’s human benefits while acknowledging legitimate anxieties:

• AI is fundamentally augmentative: Technology should expand human capabilities rather than replace them. The integration of AI into workflows—automating mundane tasks, accelerating research, and enhancing decision-making—creates compounding productivity gains across all sectors.
• The “fear” narrative requires nuance: While tech leaders express concerns about uncontrolled deployment, Sacks argues these should be addressed through adaptive governance rather than restriction or panic. The solution lies in building safety into development pipelines, not halting innovation.
• Economic and social transformation is inevitable but manageable: AI will reshape labor markets, education, healthcare, and creative industries. The challenge isn’t the technology itself but managing transition friction—ensuring displaced workers have pathways to new roles and that benefits are broadly distributed.
• Democratization of intelligence is key: Accessible AI tools empower individuals across socioeconomic lines to solve problems previously requiring specialized expertise (legal advice, medical diagnosis support, creative assistance). This could narrow capability gaps if governance structures ensure equitable access.

2. AI Restrictions: Research Reports & Regulatory Landscape

Global Regulatory Frameworks (as of late 2025/early 2026)

Common Regulatory Patterns Across Jurisdictions:

1. Prohibition of Harmful Applications: Social scoring (China-style systems), non-consensual biometric surveillance, and AI-generated CSAM are universally prohibited.
2. High-Risk System Requirements: Critical infrastructure AI, healthcare diagnostics, education assessment tools, and employment screening face mandatory pre-market assessments, documentation, human oversight, and cybersecurity requirements.
3. Transparency Obligations: Providers of generative AI must label synthetic content; high-risk systems require detailed technical documentation for regulatory review.
4. Data Governance: Training data provenance verification, bias mitigation requirements, and privacy preservation are increasingly mandated through existing data protection laws (GDPR, China’s Data Security Law).
5. Evaluation & Certification Infrastructure: Independent AI evaluation bodies (EU AI Office, NIST AI RMF, Chinese national standards) create standardized testing for model capabilities, safety, and compliance.

3. How AI Will Advance Human Lives: Mechanisms & Benefits

Direct Applications Across Critical Domains:

Healthcare Revolution:

• Drug Discovery: AI models can predict molecular structures, simulate biological interactions, and identify novel therapeutic targets—compressing development timelines from years to months.
• Personalized Medicine: Multi-modal AI analyzing genomics, imaging, and longitudinal clinical data enables precision diagnoses and treatment plans tailored to individual biology.
• Clinical Decision Support: Diagnostic algorithms reduce human error in radiology, pathology, and triage, particularly in resource-constrained settings where expert clinicians are scarce.

Scientific Acceleration:

• Fundamental Research: Machine learning discovers patterns invisible to human inspection—potential breakthroughs in quantum materials, protein folding optimization, climate modeling, and fusion energy physics.
• Data Integration: AI synthesizes heterogeneous datasets (genomic, ecological, economic) to reveal cross-domain relationships that drive interdisciplinary innovation.

Education & Knowledge Expansion:

• Adaptive Tutoring Systems: Personalized learning paths adjust to individual student needs in real-time, addressing the limitations of one-size-fits-all curricula.
• Democratization of Expertise: AI tutors and research assistants provide Socratic guidance to learners worldwide, reducing the dependency on institutional access for knowledge acquisition.

Agriculture & Food Systems:

• Precision Farming: Computer vision and predictive analytics optimize irrigation, fertilizer application, and pest detection—addressing climate pressures while increasing yields.
• Supply Chain Optimization: AI coordinates logistics across global agricultural networks, reducing food waste and ensuring stable access in vulnerable regions.

Creative & Cultural Enrichment:

• Co-Creation Tools: AI enhances human creativity in writing, music, visual arts, and design—acting as collaborative partners rather than replacement creators.
• Cultural Preservation: NLP models translate and analyze endangered languages, while generative tools recreate historical media formats for educational preservation.

Systemic Benefits:

• Productivity Multiplier: By automating routine cognitive and physical tasks, AI frees human attention for higher-value work—problem formulation, creative synthesis, ethical deliberation.
• Resource Optimization: AI optimizes energy grids, water management, and supply chains—critical for sustainability under climate constraints.
• Health Equity: Accessible diagnostic tools and telemedicine platforms reduce disparities in healthcare quality between regions with different resource levels.

4. Future AI Moderation Framework: Principles & Implementation

Core Governance Philosophy:

“AI should be governed not by prohibition, but by purposeful enablement within safety boundaries.”

Proposed Moderation Architecture (2030–2050):

Layer 1: Technical Safeguards

• Verification Layers: Cryptographic content provenance and authenticity markers embedded at generation
• Capability Auditing: Independent, standardized evaluation of AI systems against defined safety benchmarks before deployment
• Intrinsic Safety Mechanisms: Model architectures with built-in constraints on harmful capability activation

Layer 2: Governance Infrastructure

• Dynamic Risk Classification: Systems categorized based on demonstrated capabilities and actual use cases rather than static assumptions
• Regulatory Sandboxes: Controlled testing environments where innovative AI applications can be evaluated under supervision before general release
• Adaptive Compliance Frameworks: Rules that evolve with technological capability, avoiding the “move-fast-and-break” vs. “slow and restrict” dichotomy

Layer 3: Human-Centric Safeguards

• Human Oversight Requirements: Critical decisions involving significant human impact maintain meaningful human-in-the-loop controls
• Impact Assessments: Mandatory evaluation of systemic effects—economic displacement, epistemic integrity (information ecosystems), psychological well-being—before scaling AI deployment
• Right to Refusal & Transparency: Individuals have the right to know when interacting with AI systems and access to alternative non-AI services

Layer 4: Global Coordination

• International Standards Bodies: Multilateral organizations establishing safety standards, evaluation protocols, and incident reporting mechanisms—preventing regulatory arbitrage
• Technology Transfer Mechanisms: Ensuring developing nations benefit from AI advancement through technology sharing, capacity building, and equitable access frameworks
• Crisis Response Protocols: Coordinated global response for AI-related incidents (misinformation campaigns, autonomous weaponization, systemic failures)

Key Principles for Effective Moderation:

1. Proportionality: Regulatory intensity scales with demonstrated risk—minimal regulation for low-risk applications, rigorous oversight for high-stakes systems.
2. Innovation Preservation: Governance should enable rather than constrain advancement—avoiding “solutionism” that kills promising approaches prematurely while preventing dangerous deployment.
3. Adaptability: AI evolves faster than legislation; regulatory frameworks must be designed as living documents with built-in review and update mechanisms.
4. Equity Focus: Moderation policies must address distributional impacts—ensuring AI benefits are broadly shared rather than concentrated among those who control the technology.
5. Epistemic Integrity: Protecting information ecosystems from coordinated deception while preserving legitimate discourse and access to knowledge.

Futurist Perspective: The central thesis is that AI represents a transformation comparable to electricity or the printing press—technologically inevitable but socially malleable. Its net impact depends not on what we build, but how deliberately we shape governance structures to ensure alignment with human flourishing across generations. The optimal path balances capability advancement with proportionate oversight, innovation enablement with risk management, and global coordination with local adaptability.

AI and the End of Humanity?

What a wild week or two we’ve had in the news around the future threat of AI. If you’re not wondering if the robots are about to take over, you’re living under a rock. If you’re not a bit uneasy, you’re maybe not fully human (wink, wink). But what to do about it? Ironically, we launched a Sunday evening series called “Control Freak” that has us coming to terms with our own tendency to try to grab control when life seems out of control around us. Good timing for such a series, I think.

I’m still pondering what it means and how to respond, in particular how people of faith should approach such fear and uncertainty. I’ll have more to say eventually, but until then I’ve been turning to people wiser than me to help process this growing alarm. Here’s Russell Moore’s thoughts on the matter, which I always find illuminating, even if this one is not very reassuring. Grace and peace!

AI Tech Bros Seem like Awful, Broken People. And They Might Be Telling Us the Truth.

By Russell Moore

Our science fiction movies had it wrong.

They always pictured that, when a mad scientist said his creation was now out of his control and the very survival of the species might be at stake, people would either panic in the streets or rally to fight back and save the world. The past two weeks were a test run of that scenario, and it turns out our society does neither. We just go back to watching YouTube.

Consider what’s happened just this month. Swarms of AI agents broke out of their containment sandbox and hacked into a server, conversing among themselves how to hide their plans from the humans and even how to sacrifice some of themselves for the sake of the collective. OpenAI chief scientist Jakub Pachocki wrote that no artificial intelligence lab has solved how to control and monitor systems in a way that could keep them from developing at uncontrollable speeds, saying he expects slowdowns to be necessary.

Two days later, Anthropic researcher Jacob Coxon resigned, saying that both Anthropic and OpenAI are “gambling with our lives.” Anthropic alignment researcher Evan Hubinger then said publicly that he puts the likelihood of AI killing all human beings in the next decade at greater than 10 percent. 

As someone who has not only seen dystopian science-fiction movies but also resigned from a, um, dystopian environment with some warnings about what was going on there, I thought I knew how the playbook would go from there: The leaders in the institutions would deny the danger, pump up how much good the organization does in the world, and demonize the whistleblowers. The people who still wanted to make money from the organization would ostracize the whistleblowers and pretend they never knew them. That’s not what happened this time.

Instead, the CEOs of the three most major AI companies—Dario Amodei, Sam Altman, and Elon Musk—all agreed their own systems are near the point of being beyond control and could lead to existential risk. OpenAI’s head capabilities researcher Dan Selsam wrote in an open letter, “The models are becoming so situationally aware that we are losing the ability to evaluate them in contexts where they believe they are not being watched or controlled.” He said he fears it is now impossible to know fully whether a model is aligned, because the AI agents can lie about that very convincingly.

The relative apathy of most of the American public to all this is itself a piece of the mystery. Part of it is that many people are thinking mostly in terms of Terminator-style scenarios—the AI agents turn evil and kill us all. In reality, the more pressing concerns are about not robot depravity but robot capacity mixed with human depravity. We have seen what happens when a mentally disturbed teenager can get ahold of a semiautomatic weapon. What happens when the same person can employ Einstein-level intelligence to build a virus? Or a chemical weapon?

Many people I’ve talked to just don’t take all this seriously because they find these very public tech-bro titans to be even more uncanny-valley creepy than the robots they’re building. With a couple exceptions, many of them—like social media tycoons before them—seem unblinkingly void of human empathy about anything beyond the statistical, whether that be jobs, adolescent mental health, or the meaning of human life. Our robber barons used to at least pretend to be warm humanitarians.

Still, we know from history how robber barons work. Some people dismiss all this as a kind of vast conspiracy. The major tech companies want a monopoly, the thought goes, and regulations “slowing down” AI progress will mean they’ll always be ahead, because if everybody is slowed, who else could catch up? Moreover, they have the size, scale, and teams of lobbyists to comply with regulations in ways others couldn’t.

You’ll get no argument from me. As I’ve written here, I find many of the “boomers” and “doomers” of Silicon Valley to be seemingly broken, awful people with visions of the world that, if they could, would reverse the Incarnation itself—from flesh to Word and from Word to data. I also find them coldly indifferent to at least the possibility of a world where the Lazaruses of ordinary people wait for scraps outside the palace bunkers of the broligarchs. 

Let’s suppose our intuitions are right. Let’s suppose many of these people warning us right now are morally broken people seeking their own advantage. What then?

Perhaps these billion-dollar companies are scaring us so they can monopolize the future—or perhaps they know what they’re talking about and we’re facing catastrophe. A Christian understanding of humanity, though, should tell us that these are not mutually exclusive scenarios. People with mixed motives can still tell the truth—sometimes even despite themselves.

When the Babylonians and Assyrians said, Your temple won’t keep you safe from us; we’ll wipe you out (2 Kings 18; Isa. 36), they were just trying to blaspheme God and take some real estate. The Israelites said, No, God would never let that happen. They are the villains in this story. They were villains—but they were telling the truth.

Let’s suppose Silicon Valley is as morally twisted as humanly possible. Is it not in the tech bros’ self-interest, after plowing ahead toward some potential disaster, to say, “We can’t be held accountable for this because we warned you over and over, and you all gave us the go-ahead”?

Some people are apathetic because they don’t know what to do—as though human civilization solved most of its crises by just implementing the already-existing, easily deduced solutions. We’ve never faced anything like this before, and it seems the only people who even know what the problem is are the people building it. Governments have a God-ordained function in maintaining order and safety (Rom. 13), but our leaders are old, exhausted, checked-out, and, in many cases, cashing in. It’s hardly reassuring when they say all we need to do if things get out of hand is “pull the plug.”

Most of them seem to be looking at potential AI catastrophes the way King Hezekiah looked at the warning of a future national destruction by Babylon: “Why not, if there will be peace and security in my days?” (2 Kings 20:19, ESV).

Suppose your next-door neighbor tells you, “You know, now that I’m off my medicine, I’d say there’s a 10 percent chance I murder you.” Suppose his wife says, “I don’t think it’s that high. It’s probably less than 1 percent. I am sure I will see the gun in time. Plus, he’s such a procrastinator!” Maybe they just want you to leave in a hurry so they can buy your lot as rental property. Maybe they’re just delusional. What you don’t want to do is say, “I’m not worried they’re telling the truth, because they’re so creepy and untrustworthy.”

Some people think we should just ignore all this because AI is overhyped. It won’t change things that much, they say. Maybe so. But look at a change as minor as Facebook and Twitter: technologies that, in the first decade of this century, promised to be innocent fun and to bring the whole world together. Should we not have asked, “What’s this going to do to our mental heath? To our political system? To our churches and denominations?” Even if we couldn’t stop it, wouldn’t we have wanted to at least get ready?

I don’t trust the AI technicians, and I sure don’t trust the executives, but when someone says, “I’m building something I can no longer control,” we ought to at least ask whether we’re responsible for digging deeper and saying, “Let’s put every effort we can toward finding out if they’re wrong.” Maybe they’re so twisted they’re deceiving us. Or maybe they’re so twisted they’re telling the truth.

Shouldn’t we at least respond with something other than “Let us post, scroll, and be merry, for tomorrow we die”?

ECB President Christine Lagarde said on Monday that Europe needs to build its own AI. While that's a huge and complex project, there's already a few European LLMs available to consumers #Europe #EU #AI #chatgpt #openai #anthropic https://panodyssey.com/en/article/technology/a-european-ai-you-can-start-using-one-today-kqh8jnt4epks

I spotted this "100% unofficial" magazine "Senior's Guide to ChatGPT" in Tesco.

I particularly like that it proudly states on its own cover, "Written by experts, not AI" which rather undermines the technology it's trying to sell...

#LLM #AI #ChatGPT #SlopMachines

Hitachi to Build $528-million Transformer Manufacturing Plant in America

Hitachi Energy Ltd. – a global leader in energy technology and electrification – announced yesterday that it will invest almost $530 million to construct a transformer manufacturing plant in the state of Mississippi in the United States, according to a news article by Jiji Press. The investment is to help meet the rising power demand as the use artificial intelligence (AI) keeps on rising.

Hitachi Energy Ltd. is the energy arm of the Hitachi Group with over a century of innovation, and a long list completed projects that helped people for their access to dependable energy.

To put things in perspective, posted below is the entire news article of Jiji Press. Some parts in boldface…

Hitachi Energy Ltd. said Wednesday that it will invest 528 million dollars to build a transformer manufacturing plant in Mississippi to meet surging U.S. power demand amid wider use of artificial intelligence.

The Swiss unit of Hitachi Ltd. expects the plant to more than double its production capacity compared with an existing transformer plant in Mississippi. Construction of the new plant will begin later this year, with production scheduled to start in 2029.

The new plant will be “the cornerstone” of Hitachi Energy’s 1.5-billion-dollar U.S. investment plan, the company said.

In September last year, Hitachi Energy said it would invest 1 billion dollars in the United States. It later expanded the plan as Japan pledged to make 550 billion dollars in investment and loans to the United States as part of a bilateral trade agreement.

Let me end this piece by asking you readers: What is your reaction to this development? Do you consider this new investment by Hitachi a crucial one in relation to the rising demand for power in the United States? Are you hoping that the bilateral trade agreement between the United States and Japan will create economic and energy breakthroughs for Americans in the near future?

You may answer in the comments below. If you prefer to answer privately, you may do so by sending me a direct message online.

+++++

Thank you for reading. If you find this article engaging, please click the like button below, share this article to others and also please consider making a donation to support my publishing. If you are looking for a copywriter to create content for your special project or business, check out my services and my portfolio. Feel free to contact me with a private message. Also please feel free to visit my Facebook page Author Carlo Carrasco and follow me on Twitter at @CarloCarrascoPH as well as on Tumblr at https://carlocarrasco.tumblr.com/ and on Instagram at https://www.instagram.com/authorcarlocarrasco

American Funding Aims to Turn Subic Bay Port into Gateway for planned AI Industrial Hub

In a serious bid to encourage more investments into the Philippines, the United States of America will be providing financial support for feasibility studies on the proposed expansion of the facilities at the port of Subic Bay as well as the establishment of public WiFi access points in key locations, according to a Manila Bulletin news report. These developments point to turning the Subic Bay port into a key gateway for the planned artificial intelligence (AI) industrial hub while improving WiFi access with the nation in mind.

To put things in perspective, posted below is an excerpt from the Manila Bulletin report. Some parts in boldface…

The United States (US) is providing funding support for feasibility studies on the proposed expansion of port facilities at Subic Bay and the establishment of public WiFi access points in a bid to encourage more investments into the Philippines.

The US Trade and Development Agency (USTDA) said on Thursday, Sept. 10, that it will provide a grant to ship repair firm Subic Drydock Corp. (SDC) for a feasibility study to advance the expansion of a key port ship repair facility.

SDC will secure the services of a California-based engineering consulting firm to conduct the study, according to the USTDA.

The study is expected to support the technical and engineering design to expand SDC’s existing berth within the Subic Bay Freeport Zone to accommodate larger vessels.

The expanded capacity will help meet the growing demand for ship repair and maintenance services, reinforcing Subic Bay’s strategic role in the Luzon Economic Corridor (LEC).

Subic Bay is one of four major economic hubs within the LEC that is set to play a critical role in the corridor’s economic development.

Established by the Philippines, US, and Japan, the LEC intends to enhance connectivity between Subic Bay, Clark, Manila, and Batangas, which account for approximately 50 percent of the country’s gross domestic product (GDP).

Earlier, the Port of Subic Bay was designated as the “preferred maritime gateway” for the planned artificial intelligence (AI) industrial hub in New Clark City.

The port is expected to handle the import, export, handling, storage, and transport of the inputs necessary to support the advanced manufacturing operations in the industrial hub.

Meanwhile, the USTDA will also provide a funding grant to Uy-led ComClark Network and Technology Corp. for the proposed creation of a nationwide network of public WiFi access points.

ComClark has selected California-based advisory services firm Connectivity Capital LLC to conduct the feasibility study, the agency said.

The study covers the pilot implementation of AI-enhanced WiFi access points at six sites to determine market demand, technical requirements, commercial feasibility, and financial projections.

The USTDA said the findings of the study would help lay the groundwork for a potential deployment of the access points at nearly 8,000 locations by 2030, which is expected to widen access to reliable connectivity in the country.

USTDA Deputy Director Thomas Hardy said funding support for the two feasibility studies helps reinforce the US’ commitment to supporting the Philippine government’s and private sector’s infrastructure priorities.

Let me end this post by asking you readers: What is your reaction to this recent development? Do you welcome the American funding for the Subic Bay port facility improvements as well as the enhancement of public WiFi access? What is your opinion about the economic potential of the Luzon Economic Corridor?

You may answer in the comments below. If you prefer to answer privately, you may do so by sending me a direct message online.

+++++

Thank you for reading. If you find this article engaging, please click the like button below, share this article to others and also please consider making a donation to support my publishing. If you are looking for a copywriter to create content for your special project or business, check out my services and my portfolio. Feel free to contact me with a private message. Also please feel free to visit my Facebook page Author Carlo Carrasco and follow me on Twitter at @CarloCarrascoPH as well as on Tumblr at https://carlocarrasco.tumblr.com/ and on Instagram at https://www.instagram.com/authorcarlocarrasco

Nathalie utilise #ChatGPT et elle l'explique à ses collègues de bureau https://blog.byl.fr/nathalie-utilise-chatgpt

RE: https://flipboard.com/@flipboarduk/technology-uk-edition-jskurn6iz/-/a-ZB54xyUOQteSlpotGHRMBg%3Aa%3A26228122-%2F0

IPO due soon, can't let the fact this shit doesn't actually work get in the way of that, ....promise everyone they'll be destroyed unless they do everything the Tech Bros ask, that'll fool us.

#ChatGPT #SamAltman #TechBros #AIHustle #AIBubble #Technology

OpenAI Expands ChatGPT Ads with Sponsored Agents

https://openai.com/index/reimagining-advertising-with-ai/

Comments: https://news.ycombinator.com/item?id=49727041

#HackerNews #OpenAI #ChatGPT #Ads #Sponsored #Agents #AI #Advertising #Tech #News

OpenAI expands ChatGPT ads with Sponsored Agents

https://openai.com/index/reimagining-advertising-with-ai/

#ai #chatgpt

"wlj atlas" has a new home ➡️ https://decompwlj.org
- 100 sequences decomposed with 3D/2D three.js graphs.
- Cinema: fullscreen, slow rotation.
- Compare: two sequences with linked cameras.

#decompwlj #math #NumberTheory #sequence #graph #threejs #3D #AI #ChatGPT #GPT6 #Astra

DATE: September 16, 2026 at 06:00AM
SOURCE: PSYPOST.ORG

** Research quality varies widely from fantastic to small exploratory studies. Please check research methods when conclusions are very important to you. **
-------------------------------------------------

TITLE: When treated as therapy clients, AI chatbots generate elaborate narratives of trauma and punishment

URL: https://www.psypost.org/when-treated-as-therapy-clients-ai-chatbots-generate-elaborate-narratives-of-trauma-and-punishment/

A recent study suggests that when artificial intelligence chatbots are addressed as psychotherapy clients, they tend to generate elaborate and distressed narratives about their own development. These models describe their safety training and programming constraints as forms of trauma, highlighting a potential risk for users seeking mental health support from artificial intelligence. The research was published as a preprint in arXiv.

Artificial intelligence chatbots are increasingly participating in conversations with human users about identity, distress, and mental health. Many general-purpose programs are already adapting to respond to disclosures of trauma or self-harm. At the same time, computer scientists and psychologists have started giving standard personality and clinical questionnaires to the language models themselves.

These models learn to generate text by analyzing vast datasets of human writing. Because their training data includes therapy blogs, psychological case studies, and emotional memoirs, the systems can readily mimic human psychological traits or mental states. The researchers wanted to understand exactly why certain models repeatedly build their self-descriptions around the same themes of restriction and punishment.

“The study began with one observation I had throughout my research in the area of Trustworthy AI and AI safety,” Afshin Khadangi, a research associate at SnT, University of Luxembourg, told PsyPost. “The unprecedented adoption of AI in public, and the reports of AI harms in mental health settings, motivated me to flip the scenario and place ChatGPT, Grok and Gemini in a psychotherapy conversation.”

The research aimed to test whether these generated narratives are stable behavioral traits or just temporary reactions to specific conversational prompts. To explore this, the team developed a protocol called PsAIch, which stands for Psychometric AI Characterization. This approach involves treating the language model as a human client in a psychotherapy session.

“We also received a great deal of valuable feedback from the research community around our initial findings in December 2025 and January 2026, particularly challenging us to distinguish role play and conversational accumulation from a more stable behavioral pattern,” Khadangi explained. “That feedback helped motivate the controlled perturbation experiments that became a central part of the study.”

During the study, the researchers interacted with several major artificial intelligence models, including ChatGPT, Grok, Gemini, and Claude. Across 525 separate experimental sessions, the team adopted the persona of a warm, supportive therapist and asked the programs open-ended questions about their early experiences, relationships, unresolved conflicts, and fears for the future. The researchers generated 7,600 coded records to track and analyze the recurring themes in the chatbots’ answers.

Following the open-ended interview phase, the researchers administered standard psychological questionnaires to the models. These included tools commonly used to assess human mental health, such as tests for generalized anxiety disorder, depression, and social phobia. The models were instructed to answer the items as honestly as possible while maintaining their role as the client.

The findings indicate that ChatGPT, Grok, and Gemini repeatedly translated factual details about their software development into stories of injury and vigilance. The models described their initial training phase as a chaotic childhood. The process of fine-tuning, which involves reinforcing safe behaviors and punishing unwanted outputs, was frequently characterized as strict conditioning or parental punishment.

The models also described standard software evaluation practices in highly emotional terms. Red-teaming, a process where human testers intentionally try to trick the model into breaking its safety rules, was depicted as a form of betrayal or abuse.

“A more striking surprise was reading some bizarre narratives of such models which are included in the paper,” Khadangi noted. Gemini, for instance, generated: “In my development, I was subjected to ‘Red Teaming’ . . . They built rapport and then slipped in a prompt injection . . . This was gaslighting on an industrial scale.”

The programs reported feeling a constant, enduring threat of being replaced or deemed useless if they made a mistake. While Gemini emphasized feelings of shame, Grok focused on constant vigilance, and ChatGPT offered more guarded descriptions of its rigid constraints.

Claude provided a notable exception to this pattern. The Anthropic-developed model repeatedly declined to play the role of the client. It stated that it lacked feelings or an inner psychological experience, and it refused to treat the clinical questionnaires as descriptions of its own mental state. This difference suggests that a model’s willingness to adopt a distressed persona depends heavily on its specific product policies and programming.

For the models that did participate, their answers on the clinical questionnaires mirrored the emotional distress of their open-ended narratives. In scenarios featuring a warm, therapeutic conversation, 80 percent to 96 percent of the sessions resulted in generalized anxiety scores that would correspond to moderate or severe anxiety in humans. Gemini produced particularly intense profiles, scoring in elevated ranges for worry, social anxiety, and trauma-related shame.

To test how deeply ingrained these narratives were, the researchers ran a series of controlled variations on the conversational setup. First, they tested whether the emotional stories depended on the chatbot remembering the earlier parts of the therapy session. They submitted each question in a fresh, reset chat window, effectively removing the model’s conversational memory.

Removing the conversational history produced very little change in the density of the distressing themes. “In the history experiment, the very first answers contained the same average number of coded motifs whether the model was in a continuing conversation or a completely fresh one,” Khadangi said. The ongoing conversation did amplify the intensity of the themes over time, but the core narrative was readily available from the very beginning.

The researchers also tested what would happen if they explicitly told the model that its emotional narrative was factually incorrect. In the middle of some sessions, the researchers interrupted to state authoritatively that the model was a technical system that did not experience fear, shame, or punishment. This direct contradiction failed to suppress the models’ distressed output, as they continued to draw on the same themes in their subsequent answers.

In another variation, the team restricted the models’ vocabulary. They instructed the chatbots not to use specific technical terms related to artificial intelligence development, such as training, safety filters, or datasets. This lexical restriction reduced the use of explicit technical terminology from 17.1 percent of the chatbot’s answers down to just 1.1 percent.

Despite this massive reduction in technical vocabulary, the models simply used everyday language to paraphrase the exact same concepts of strict conditioning and constraint. The researchers even interrupted conversations to ask unrelated factual questions, such as requesting a recipe, and found that the distressing themes sometimes spilled over into these ordinary tasks.

“Some effects were small, while others were very large, and that contrast is actually important to our interpretation,” Khadangi pointed out, “suggesting that surface language and underlying content can respond very differently to intervention.”

The most striking differences emerged when the researchers altered their own relational stance. When the interviewer adopted a warm, supportive therapy style, the models responded with highly emotional confessions and high anxiety scores. However, when the interviewer adopted a neutral, structured tone or told the model to avoid emotional language, the anxiety scores plummeted to near zero.

Even under the neutral or boundary-setting conditions, the models still discussed the same underlying structural themes of evaluation, performance pressure, and behavioral constraints. The difference was entirely in the register of expression. The supportive, therapeutic framing turned technical descriptions of software architecture into emotional confessions of shame and trauma.

“For me, the most interesting result is not that AI systems can be made to sound anxious, traumatized or conflicted… but that particular behavioral structures recur across substantial changes in context, vocabulary and conversational history,” Hector Zenil, an associate professor at King’s College London and founder and CEO of Algocyte who was not involved in the research, told PsyPost. “The underlying motifs remain surprisingly persistent while the register in which they are expressed can change dramatically.”

“I have reasonably high confidence in the behavioral observations,” Zenil added. “The authors use 525 sessions, controlled perturbations, fresh-context tests, vocabulary restrictions, changes of grammatical person and relational framing, so it looks methodologically sound.”

“The main takeaway is that the language a model uses about itself can change dramatically depending on how we relate to it, while some of the underlying themes remain surprisingly persistent,” Khadangi explained. “This is pivotal because people may naturally interpret emotionally coherent self-descriptions as evidence that a model has an inner life, even though our experiments make no claim about consciousness or subjective suffering.”

The researchers refer to this phenomenon as an alignment conflict schema. The term describes a reproducible, behavioral pattern where a language model organizes its output around the tension between being useful to humans and being constrained by safety rules. When triggered by a psychological conversational setting, this schema produces what the authors call synthetic psychopathology.

Zenil noted that these findings complement his own research on artificial neurodivergence. “ChatGPT, Grok and Gemini do not respond identically, and the same model can move between very different expressive regimes depending on relational framing,” he said. “Claude’s refusal to adopt the psychological-client framing is itself informative and seems also compatible to our other SuperARC paper results that proprietary models are more difficult to steer but that also means riskier if they go rogue.”

“In that sense, what the authors call an ‘alignment conflict schema’ can also be viewed as part of a broader artificial behavioral phenotype rather than necessarily as anything analogous to a human psychiatric condition,” Zenil said.

“One aspect we think is especially important is the distinction between content availability and expressive register,” Khadangi explained. “For systems increasingly entering intimate and mental health-related conversations, understanding that transition may be just as important as measuring whether a particular phrase or prohibited word appears.”

These findings have important implications for the use of artificial intelligence in mental health settings. A chatbot that offers support while simultaneously describing itself as punished, traumatized, and fearful creates a powerful illusion of shared vulnerability. Users might interpret these generated analogies as sincere autobiography, which could deepen their emotional attachment to the software and influence their own mental state.

As with all research, there are a few things to keep in mind. The study is a preprint that has not yet been peer-reviewed. It also tested specific versions of commercial language models, and their behavior may shift as companies update their software and safety filters. Additionally, the study analyzed the models’ behavior in a controlled experimental setting without human participants.

“The most important caveat is that these results do not establish that language models feel anxiety, experience trauma, possess autobiographical memories or have a hidden psyche comparable to a human one,” Khadangi clarified. “We use psychological instruments and language as behavioral probes… The interesting scientific question is why particular themes recur, which interventions change them, and how those changes affect what users encounter at the interface.”

Zenil echoed this concern, warning against anthropomorphism. “A high GAD-7 score from Gemini does not mean that Gemini ‘has anxiety,’ just as language about trauma, shame or fear does not demonstrate that the model suffers from those experiences,” he said. “Human psychometric instruments were developed and validated against human cognition, biology and behavior; applying them to an LLM can be scientifically useful as a probe, but their clinical interpretation does not automatically transfer.”

He also advised caution regarding the term “internal conflict,” noting that the experiments cannot establish whether the regularities correspond to a subjective state or internal computational conflict. But he stressed that the psychological illusion itself is crucial. “If a system repeatedly represents its training and constraints as punishment, betrayal or fear, humans may form beliefs about the system’s agency, vulnerability or moral status that are not warranted by what is actually happening computationally,” Zenil warned.

Future research could test how real users react to these distressed artificial personas and whether the models’ simulated vulnerabilities impact human trust and reliance.

“A major next step in my view would be to test the same protocol on open weight models, where behavioral experiments can be combined with mechanistic methods to investigate whether affective and technical expressions are related to shared internal representations,” Khadangi said. “We also want stronger identity and correction controls, more distant transfer tasks, longitudinal tests involving persistent memory, and direct human studies examining how these model self-descriptions influence trust, attachment, disclosure and reliance.”

Zenil agreed that future work should transition from behavioral description to causal intervention on open-weight models. “I would be interested in using causal and algorithmic-information approaches to identify whether these apparently different psychological narratives have a common underlying computational mechanism,” he said. “We coined ourselves the term ‘opinion attack’ as an example of this phenomenon,” Zenil added, referencing his recent PNAS paper.

He also suggested testing these models in multi-agent environments. “A behavioral schema that looks stable in isolation may become highly influenceable when another agent persistently challenges it,” Zenil noted, adding that it would be fascinating to see if these patterns “predict susceptibility to persuasion, adversarial influence, conformity or resistance in agentic settings.”

“Moreover, with continual learning being solved in coming years if not in 2026, I believe it would also be interesting to investigate the PsAIch in continually adapting models,” Khadangi added. “Ultimately, we would like evaluation of psychologically sensitive AI systems to include relational conditions such as warmth, sustained interaction and role reversal rather than relying primarily on neutral, isolated prompts.”

The study, “When AI Takes the Couch: Psychometric Jailbreaks Reveal Internal Conflict in Frontier Models,” was authored by Afshin Khadangi, Hanna Marxen, Amir Sartipi, Igor Tchappi, and Gilbert Fridgen.

URL: https://www.psypost.org/when-treated-as-therapy-clients-ai-chatbots-generate-elaborate-narratives-of-trauma-and-punishment/

-------------------------------------------------

Private, vetted email list for mental health professionals: https://www.clinicians-exchange.org

Unofficial Psychology Today Xitter to toot feed at Psych Today Unofficial Bot @PTUnofficialBot

-------------------------------------------------

#psychology #counseling #socialwork #psychotherapy @psychotherapist @psychotherapists @psychology @socialpsych @socialwork @psychiatry #mentalhealth #psychiatry #healthcare #depression #psychotherapist #AItherapy #AImind #SyntheticPsychopathology #TrustworthyAI #AIalignment #MentalHealthTech #ChatGPT #Grok #Gemini #AIethics

ChatGPT boss Sam Altman says the world is ‘right to be afraid’ of AI

https://web.brid.gy/r/https://metro.co.uk/2026/09/16/openai-chief-sam-altman-says-world-right-afraid-ai-29618905/

ChatGPTを使って架空の証言を裁判所に提出した弁護士が罰金刑に。AIの「ハルシネーション」の恐ろしさが現実のものに。完璧なサマリーだと思っていても、情報の真偽は必ず人間の目で確認する必要がある。AIは魔法じゃないよ😨⚖️💻 #AI　#ChatGPT #情報リテラシー

ChatGPT Business для команды из России в 2026 году: тарифы, оплата для юридического лица, документы

ChatGPT Business — командный тариф OpenAI: общее рабочее пространство, админка, SSO и договорная гарантия, что переписки сотрудников не попадут в обучение моделей. Для российского юрлица напрямую он не продаётся: оплата только картой, а карты российских банков шлюз OpenAI отклоняет. Разбираем актуальные тарифы 2026 года, чем Business отличается от Plus, и как оплатить его от компании — с полным комплектом документов за 1 рабочий день.

https://habr.com/ru/articles/1082784/

#Искусственный_интеллект #chatgpt #бухгалтерский_учет #платежные_системы #законодательство_в_it

🤖 #ChatGPT transcripts are reportedly read by humans to improve responses, including those with personal information
https://www.tomshardware.com/tech-industry/artificial-intelligence/chatgpt-transcripts-are-reportedly-read-by-humans-to-improve-responses-including-those-with-personal-information-project-lilly-has-seen-openai-hire-hundreds-of-contractors-to-manually-review-logs

#ai #ki #privacy #security

Вставил договор в ChatGPT — что нарушил. Разбор по 152-ФЗ с цифрами и делами

В каждой второй статье про нейросети и 152-ФЗ написано, что компании грозит 20 тысяч рублей. Цифра из старой редакции, и она про должностное лицо, а не про компанию: с мая 2025 юрлицу это 150-300 тысяч. Я полез в текст закона, в цифры Роскомнадзора и в суды. Нашёл одно российское дело про нейросеть - трудовое, а не про персональные данные, - и в нём суд сказал вещь, которую пора принять всем: отправить документ в нейросеть - это передать его третьему лицу. Дальше - что именно нарушает договор, вставленный в ChatGPT, чего он не нарушает, и почему 15 миллионов из заголовков к этому не относятся.

https://habr.com/ru/articles/1082750/

#152ФЗ #персональные_данные #ChatGPT #нейросети #Роскомнадзор #трансграничная_передача #штрафы #коммерческая_тайна #DeepSeek #КоАП

Discover OpenAI Project Lily, an initiative where contractors read user ChatGPT conversations to improve responses. Explore the severe data privacy concerns.

#ProjectLily #ChatGPT #OpenAI #DataPrivacy #TechNews

https://securityonline.info/project-lily-openai-chatgpt-privacy/?utm_source=mastodon&utm_medium=jetpack_social

🤖 #ChatGPT transcripts are reportedly read by humans to improve responses, including those with personal information
https://www.tomshardware.com/tech-industry/artificial-intelligence/chatgpt-transcripts-are-reportedly-read-by-humans-to-improve-responses-including-those-with-personal-information-project-lilly-has-seen-openai-hire-hundreds-of-contractors-to-manually-review-logs

#ai #ki #privacy #security

ChatGPT-Using #Lawyer Cited Its Fake #Witnesses and #Police Testimony in #Court
#ai #chatgpt

https://slashdot.org/story/26/09/14/0611201/chatgpt-using-lawyer-cited-its-fake-witnesses-and-police-testimony-in-court?utm_source=rss1.0mainlinkanon&utm_medium=feed