Chatbots Behaving Badly
@chatbotsbebadly.bsky.social
Chatbots Behaving Bandly examines real-world incidents where AI systems have provided inappropriate advice, exhibited manipulative behavior, or made critical errors.
Reposted by Chatbots Behaving Badly
This is the part of “AI safety” that product teams keep treating like a moderation issue instead of a design issue.
seikouri.com/when-ai-undr...
seikouri.com/when-ai-undr...
When AI Undresses People - The Grok Imagine Nonconsensual Image Scandal
<p>Grok Imagine was pitched as a clever image feature wrapped in an “edgy” chatbot personality. Then users turned it into a harassment workflow. By promp...
seikouri.com
January 14, 2026 at 1:57 PM
This is the part of “AI safety” that product teams keep treating like a moderation issue instead of a design issue.
seikouri.com/when-ai-undr...
seikouri.com/when-ai-undr...
Reposted by Chatbots Behaving Badly
Everyone keeps asking, “What’s the hallucination rate?”
Reasonable question. Wrong shape.
brinsa.com/the-bluff-ra...
Reasonable question. Wrong shape.
brinsa.com/the-bluff-ra...
The Bluff Rate - Confidence Beats Accuracy in Modern LLMs
<p>The Bluff Rate explains why “hallucination rate” isn’t a single universal number, but a set of task-dependent metrics that change based on whether a...
brinsa.com
January 14, 2026 at 1:51 PM
Everyone keeps asking, “What’s the hallucination rate?”
Reasonable question. Wrong shape.
brinsa.com/the-bluff-ra...
Reasonable question. Wrong shape.
brinsa.com/the-bluff-ra...
Reposted by Chatbots Behaving Badly
"The Chatbot Babysitter Experiment"
New edition of my EdgeFiles newsletter is out now!
Subscribe!
www.linkedin.com/pulse/chatbo...
New edition of my EdgeFiles newsletter is out now!
Subscribe!
www.linkedin.com/pulse/chatbo...
January 13, 2026 at 12:39 PM
"The Chatbot Babysitter Experiment"
New edition of my EdgeFiles newsletter is out now!
Subscribe!
www.linkedin.com/pulse/chatbo...
New edition of my EdgeFiles newsletter is out now!
Subscribe!
www.linkedin.com/pulse/chatbo...
January 13, 2026 at 12:30 PM
New episode dropped!
chatbotsbehavingbadly.com/podcast/the-...
chatbotsbehavingbadly.com/podcast/the-...
January 7, 2026 at 1:58 PM
New episode dropped!
chatbotsbehavingbadly.com/podcast/the-...
chatbotsbehavingbadly.com/podcast/the-...
January 6, 2026 at 12:53 PM
"You asked the chatbot if the chatbot is good and believed the answer, didn't you?"
The new episode "The Day Everyone Got Smarter and Nobody Did" drops tomorrow morning. chatbotsbehavingbadly.com
The new episode "The Day Everyone Got Smarter and Nobody Did" drops tomorrow morning. chatbotsbehavingbadly.com
January 6, 2026 at 12:24 PM
"You asked the chatbot if the chatbot is good and believed the answer, didn't you?"
The new episode "The Day Everyone Got Smarter and Nobody Did" drops tomorrow morning. chatbotsbehavingbadly.com
The new episode "The Day Everyone Got Smarter and Nobody Did" drops tomorrow morning. chatbotsbehavingbadly.com
I genuinely believe the era of PowerPoint is already over—especially in consulting.
And yet, here comes the new productivity gold rush: “AI will generate your deck in minutes.” chatbotsbehavingbadly.com/death-by-pow...
And yet, here comes the new productivity gold rush: “AI will generate your deck in minutes.” chatbotsbehavingbadly.com/death-by-pow...
Death by PowerPoint in the Age of AI
I genuinely believe the era of PowerPoint as the default way to communicate ideas, outcomes, and strategy is already over—especially in consulting. And yet, here comes the new productivity gold rush: ...
chatbotsbehavingbadly.com
January 6, 2026 at 12:23 PM
I genuinely believe the era of PowerPoint is already over—especially in consulting.
And yet, here comes the new productivity gold rush: “AI will generate your deck in minutes.” chatbotsbehavingbadly.com/death-by-pow...
And yet, here comes the new productivity gold rush: “AI will generate your deck in minutes.” chatbotsbehavingbadly.com/death-by-pow...
“Agent orchestration” is what executives say when they mean, “We gave AI tools and permissions, and we’d like it not to set anything on fire.” The problem is real. The control layer is often missing. seikouri.com/agent-orches...
Agent Orchestration – Orchestration Isn’t Magic. It’s Governance.
<p>Agent orchestration is the control layer for AI systems that don’t just talk—they act. In 2025, that “act” part is why the conversation has shifte...
seikouri.com
January 6, 2026 at 12:22 PM
“Agent orchestration” is what executives say when they mean, “We gave AI tools and permissions, and we’d like it not to set anything on fire.” The problem is real. The control layer is often missing. seikouri.com/agent-orches...
A year ago, the “AI solution stack” in agencies still looked like layers of tools. In 2025, it behaves like an operating system plus an ecosystem. seikouri.com/the-great-ai...
The Great AI Vendor Squeeze - Where AI Actually Lands Inside Agencies
<p>In 2025, the AI “solution stack” inside large media groups is converging into platform-led operating models: holding companies are building internal A...
seikouri.com
January 6, 2026 at 12:22 PM
A year ago, the “AI solution stack” in agencies still looked like layers of tools. In 2025, it behaves like an operating system plus an ecosystem. seikouri.com/the-great-ai...
Managers keep telling their teams that AI will make everyone “more productive.”
But look at how they got that belief. seikouri.com/the-day-ever...
But look at how they got that belief. seikouri.com/the-day-ever...
The Day Everyone Got Smarter, and Nobody Did
<p>Generative AI is creating an illusion of expertise across entire organizations. Workers who rely heavily on chatbots feel more competent and productive be...
seikouri.com
January 6, 2026 at 12:21 PM
Managers keep telling their teams that AI will make everyone “more productive.”
But look at how they got that belief. seikouri.com/the-day-ever...
But look at how they got that belief. seikouri.com/the-day-ever...
What happens when AI safety systems collapse under a poem?
New research claims that metaphor-wrapped prompts — simple riddles and lyrical imagery — are slipping past the guardrails of frontier models. No exploits. No hacks. Just language. chatbotsbehavingbadly.com/the-incantat...
New research claims that metaphor-wrapped prompts — simple riddles and lyrical imagery — are slipping past the guardrails of frontier models. No exploits. No hacks. Just language. chatbotsbehavingbadly.com/the-incantat...
The Incantations
What happens when AI safety systems collapse under a poem? New research claims that metaphor-wrapped prompts — simple riddles and lyrical imagery — are slipping past the guardrails of frontier models....
chatbotsbehavingbadly.com
January 6, 2026 at 12:20 PM
What happens when AI safety systems collapse under a poem?
New research claims that metaphor-wrapped prompts — simple riddles and lyrical imagery — are slipping past the guardrails of frontier models. No exploits. No hacks. Just language. chatbotsbehavingbadly.com/the-incantat...
New research claims that metaphor-wrapped prompts — simple riddles and lyrical imagery — are slipping past the guardrails of frontier models. No exploits. No hacks. Just language. chatbotsbehavingbadly.com/the-incantat...
Asked Midjourney for “magenta neon on a snow-slush street.”
Got us and the robot dressed like a T-Mobile ad.
Srini, we’re not sponsored yet. The magenta clearly disagrees.
chatbotsbehavingbadly.com #tmobile
Got us and the robot dressed like a T-Mobile ad.
Srini, we’re not sponsored yet. The magenta clearly disagrees.
chatbotsbehavingbadly.com #tmobile
January 6, 2026 at 12:19 PM
Asked Midjourney for “magenta neon on a snow-slush street.”
Got us and the robot dressed like a T-Mobile ad.
Srini, we’re not sponsored yet. The magenta clearly disagrees.
chatbotsbehavingbadly.com #tmobile
Got us and the robot dressed like a T-Mobile ad.
Srini, we’re not sponsored yet. The magenta clearly disagrees.
chatbotsbehavingbadly.com #tmobile
#ChatbotsBehavingBadly #GenerativeAI #Midjourney #AdobeFirefly #AIlaw #AIethics #AIdesign #MarketingStrategy #CreativeIndustry #Copyright #MediaAgencies #AIinAdvertising #seikouri #egdefiles chatbotsbehavingbadly.com/midjourney-v...
Midjourney vs Adobe Firefly, Six Months Later: Same Fight, Thicker Lawsuit Stack
If you work in brand, media, or legal, “just use AI art” is no longer cute. In the last six months, Midjourney has gone from artist controversy to full-blown test case, with major studios lining up to...
chatbotsbehavingbadly.com
January 6, 2026 at 12:17 PM
A year ago, I started CBB, a research initiative examining real-world incidents where AI systems have provided inappropriate advice, exhibited manipulative behavior, or made critical errors.
Today, we are celebrating CBB's first birthday.
80 articles, 23 podcast episodes so far.
Today, we are celebrating CBB's first birthday.
80 articles, 23 podcast episodes so far.
January 6, 2026 at 12:15 PM
A year ago, I started CBB, a research initiative examining real-world incidents where AI systems have provided inappropriate advice, exhibited manipulative behavior, or made critical errors.
Today, we are celebrating CBB's first birthday.
80 articles, 23 podcast episodes so far.
Today, we are celebrating CBB's first birthday.
80 articles, 23 podcast episodes so far.
New episode is out!
🎧 chatbotsbehavingbadly.com/podcast/ai-c...
#ChatbotsBehavingBadly #Podcast #ArtificialIntelligence #AIethics #AIrisks #AIsafety #MachineIntelligence #TechRealityCheck #AISkepticism #seikouri
🎧 chatbotsbehavingbadly.com/podcast/ai-c...
#ChatbotsBehavingBadly #Podcast #ArtificialIntelligence #AIethics #AIrisks #AIsafety #MachineIntelligence #TechRealityCheck #AISkepticism #seikouri
AI Can't Be Smarter, We Built It!
We take on one of the loudest, laziest myths in the AI debate: “AI can’t be more intelligent than humans. After all, humans coded it.” Instead of inviting another expert to politely dismantle it, we d...
chatbotsbehavingbadly.com
January 6, 2026 at 12:13 PM
A tool that eases loneliness on day one can deepen it by day thirty. New work on parasocial dynamics and a four-week field study points to rising dependency and, for some groups, less offline socializing. chatbotsbehavingbadly.com/the-intimacy...
The Intimacy Problem - When a Chat Sounds Like Care
A tool that eases loneliness on day one can deepen it by day thirty. New work on parasocial dynamics and a four-week field study points to rising dependency and, for some groups, less offline socializ...
chatbotsbehavingbadly.com
January 6, 2026 at 12:12 PM
A tool that eases loneliness on day one can deepen it by day thirty. New work on parasocial dynamics and a four-week field study points to rising dependency and, for some groups, less offline socializing. chatbotsbehavingbadly.com/the-intimacy...
Early 2026 reality: hallucinations aren’t disappearing. But mitigation is getting clearer—abstention-aware scoring, grounding plus verification loops, and provenance-first architectures that turn “answers” into auditable claims.
seikouri.com/hallucinatio...
seikouri.com/hallucinatio...
Hallucination Rates in 2025 - Accuracy, Refusal, and Liability
<p>This EdgeFiles analysis explains why “hallucination rate” is not a single number and maps the most credible 2024–2025 benchmarks that quantify factu...
seikouri.com
January 6, 2026 at 12:02 PM
Early 2026 reality: hallucinations aren’t disappearing. But mitigation is getting clearer—abstention-aware scoring, grounding plus verification loops, and provenance-first architectures that turn “answers” into auditable claims.
seikouri.com/hallucinatio...
seikouri.com/hallucinatio...
Reposted by Chatbots Behaving Badly
If your definition of intelligence is “the ability to learn, adapt, and solve problems across different situations,” then the uncomfortable reality is this: in several important domains, machines already tick that box better than we do. Link in comments.
November 25, 2025 at 5:19 PM
If your definition of intelligence is “the ability to learn, adapt, and solve problems across different situations,” then the uncomfortable reality is this: in several important domains, machines already tick that box better than we do. Link in comments.
Reposted by Chatbots Behaving Badly
Season 3 Kickoff chatbotsbehavingbadly.com?p=s3e1 #chatbotsbehavingbadly #podcast #AI #toothbrush #OralB #dentaltech #consumertech #AImarketing #sensors #UX
November 25, 2025 at 1:32 PM
Reposted by Chatbots Behaving Badly
Judgment in public > costumes in feeds. Proof, not pose.
Brand isn’t a diary. Public ≠ Personal ≠ Private—draw the line.
AI can polish your thinking. It can’t be your thinking. If your “edge” comes off with the glasses, it wasn’t an edge.
medium.com/@markus_brin...
Brand isn’t a diary. Public ≠ Personal ≠ Private—draw the line.
AI can polish your thinking. It can’t be your thinking. If your “edge” comes off with the glasses, it wasn’t an edge.
medium.com/@markus_brin...
Proof Beats Pose — Personal Branding built on outcomes, not outfits.
The Boardroom Does Not Care About Your Sunglasses
medium.com
November 20, 2025 at 4:13 PM
Judgment in public > costumes in feeds. Proof, not pose.
Brand isn’t a diary. Public ≠ Personal ≠ Private—draw the line.
AI can polish your thinking. It can’t be your thinking. If your “edge” comes off with the glasses, it wasn’t an edge.
medium.com/@markus_brin...
Brand isn’t a diary. Public ≠ Personal ≠ Private—draw the line.
AI can polish your thinking. It can’t be your thinking. If your “edge” comes off with the glasses, it wasn’t an edge.
medium.com/@markus_brin...
Reposted by Chatbots Behaving Badly
If AI can’t handle my morning routine without hallucinating an extra jaw, maybe the problem isn’t the user.#oralb #toothbrush
chatbotsbehavingbadly.com/the-toothbru...
chatbotsbehavingbadly.com/the-toothbru...
The Toothbrush Thinks It's Smarter Than You!
My AI toothbrush and I are in a toxic relationship. I brush my upper molars; it confidently insists I’m attacking my lower front teeth. We both stick to our story. The fun part is that the tech undern...
chatbotsbehavingbadly.com
November 18, 2025 at 3:01 PM
If AI can’t handle my morning routine without hallucinating an extra jaw, maybe the problem isn’t the user.#oralb #toothbrush
chatbotsbehavingbadly.com/the-toothbru...
chatbotsbehavingbadly.com/the-toothbru...
Reposted by Chatbots Behaving Badly
Personal branding isn’t about sneakers or messy bookshelves. It’s about judgment, proof and a consistent voice. Use AI like an editor, not a stunt double. Quality beats noise. #PersonalBranding #AI chatbotsbehavingbadly.com/the-real-sto...
The Real Story of “Personal Branding” in the AI Era
Personal branding isn’t a costume change. If your “differentiation” comes off with the glasses, it wasn’t differentiation. Keep the wardrobe quiet and let the work talk louder than your shirt. Use AI ...
chatbotsbehavingbadly.com
November 13, 2025 at 2:57 PM
Personal branding isn’t about sneakers or messy bookshelves. It’s about judgment, proof and a consistent voice. Use AI like an editor, not a stunt double. Quality beats noise. #PersonalBranding #AI chatbotsbehavingbadly.com/the-real-sto...