Exclusive: How Claude became an emotional support bot
People who talk to Anthropic's Claude chatbot about emotional issues tend to grow more positive as the conversation unfolds, according to new Anthropic research shared exclusively with Axios.
Why it matters: Having a trusted confidant available 24/7 can make people feel less alone, but chatbots weren't designed for emotional support.
Bots have displayed troubling tendencies, like reinforcing delusional behavior or encouraging self-harm, that are especially problematic for young people or adults struggling with their mental health.
Driving the news: Anthropic released new research Thursday that explores how users turn to its chatbot for support and connection and what happens when they do.
While anecdotes of users prompting AI bots like Claude and ChatGPT for emotional support are widespread, Anthropic's report is the first formal acknowledgment by the AI provider of this use.
What they're saying:"We find that when people come to Claude for interpersonal advice, they're often navigating transitional moments — figuring out their next career move, working through personal growth, or untangling romantic relationships," per the report.
The report calls these interactions with chatbots "affective use," defined roughly as personal exchanges with Claude motivated by emotional or psychological needs.
Zoom in: The report found evidence that users don't necessarily turn to chatbots deliberately looking for love or companionship, but some conversations evolve that way.
"We also noticed that in longer conversations, counseling or coaching conversations occasionally morph into companionship — despite that not being the original reason someone reached out," per the report.
"As these conversations with Claude progress, we found that the person's expressed sentiment often becomes more positive," Anthropic societal impacts researcher Miles McCain told Axios. "And while we can't claim that these shifts represent lasting emotional benefits, the absence of clear negative spirals is reassuring."
Researchers behind the report told Axios that the results are preliminary and that measuring "expressed sentiment" can be limited.
By the numbers: Anthropic found that AI companionship isn't fully replacing the real thing anytime soon. Most people still use Claude for work tasks and content creation.
A relatively small number (2.9%) of interactions with Claude constituted "affective use" — a finding that confirms previous research from OpenAI.
Companionship and roleplay combined were 0.5% of conversations.
Romantic or sexual roleplay — which Claude's training actively discourages — was less than 0.1%, according to the report.
What they did: Anthropic analyzed user behavior with Clio, a tool it launched last year that works like Google Trends — aggregating chats while stripping out identifying details.
Clio anonymizes and aggregates Claude chats to keep specific conversations private while revealing broader trends. This is similar to the way Google tracks what people are searching for without revealing (or allowing humans to have access to) personal search histories.
The research excluded conversations that focused on writing stories, fictional dialogues, blog posts or other content.
Among conversations that included roleplaying, Anthropic says the researchers only analyzed "meaningful interactive" chats, meaning that they included four or more human messages.
Yes, but: While the internet is full of people claiming that they've cut costs on therapy by turning to a chatbot, there's plenty of evidence that bots make particularly bad therapists because they're so eager to please users.
Anthropic says it didn't study extreme usage patterns or how chatbots can reinforce delusions or conspiracy theories, which the company admits is worthy of a separate study.
To keep chats private, the researchers only looked at clusters of conversations from multiple users and didn't analyze individual users' conversations over time, which makes it difficult to study emotional dependency, Anthropic notes.
"We need large, rigorous trials that are of longer duration, because if you just relieve a person's anxiety, stress or depression on a very short term basis, that's not what we're after," physician-researcher Eric Topol told Axios. "We're after durable benefits. So I'm confident that we're going to get there, but we're not there yet."
Zoom out: Anthropic, founded by former OpenAI staff, pitches Claude as a more responsible alternative to ChatGPT.
"Safety is deeply ingrained in everything that we do, and it underpins all of our work," Alexandra Sanderford, Anthropic's head of safeguards, policy and enforcement, told Axios. "We really do try to prioritize human values and welfare."
The company has recently shared assessments of potentially alarming behavior by Claude in hypothetical test scenarios, including a willingness to blackmail users.
Last week Anthropic also released findings
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
CNET
an hour ago
- CNET
Meta Scores AI Fair Use Court Victory, but Judge Warns Such Wins Won't Always Be the Case
AI companies scored another victory in court this week. Meta on Wednesday won a motion for partial summary judgment in its favor in Kadrey v. Meta, a case brought by 13 authors alleging the company infringed on their copyright protections by illegally using their books to train its Llama AI models. The ruling comes two days after a similar victory for Claude maker Anthropic. But Judge Vince Chhabria stressed in his order that this ruling should be limited and doesn't absolve Meta of future claims from other authors. "This ruling does not stand for the proposition that Meta's use of copyrighted materials to train its language models is lawful," he wrote. "It stands only for the proposition that these plaintiffs made the wrong arguments and failed to develop a record in support of the right one." The issue at the heart of the cases is whether the AI companies' use of protected content for AI training qualifies as fair use. The fair use doctrine is a fundamental part of US copyright law that allows people to use copyrighted work without the rights holders' explicit permission, like in education and journalism. There are four key considerations when evaluating whether something is fair use. Anthropic's ruling focused on transformativeness, while Meta's focused on the effect the use of AI has on the existing publishing market. These rulings are big wins for AI companies. OpenAI, Google and others have been fighting for fair use so they don't have to enter costly and lengthy licensing agreements with content creators, much to the chagrin of content creators. For the authors bringing these cases, they may see some victories in subsequent piracy trials (for Anthropic) or new lawsuits. (Disclosure: Ziff Davis, CNET's parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.) In his analysis, Chhabria focused on the effect AI-generated books have on the existing publishing market, which he saw as the most important factor of the four needed to prove fair use. He wrote extensively about the risk that generative AI and large language models could potentially violate copyright law, and that fair use needs to be evaluated on a case-by-case basis. Some works, like autobiographies and classic literature such as The Catcher in the Rye, likely couldn't be created with AI, he wrote. However, he noted that "the market for the typical human-created romance or spy novel could be diminished substantially by the proliferation of similar AI-created works." In other words, AI slop could make human-written books seem less valuable and undercut authors' willingness and ability to create. Still, Chhabria said that the plaintiffs did not show sufficient evidence to prove harm from how "Meta's models would dilute the market for their own works." The plaintiffs focused their arguments on how Meta's AI models can reproduce exact snippets from their works and how the company's Llama models hurt their ability to license their books to AI companies. These arguments weren't as compelling in Chhabria's eyes -- he called them "clear losers" -- so he sided with Meta. That's different from the Anthropic ruling, where Judge William Alsup focused on the "exceedingly transformative" nature of the use of the plantiff's books in the results AI chatbots spit out. Chhabria wrote that while "there is no disputing" the use of copyrighted material was transformative, the more urgent question was the effect AI systems had on the ecosystem as a whole. Alsup also outlined concerns about Anthropic's methods of obtaining the books, through illegal online libraries and then by deliberating purchasing print copies to digitize for a "research library." Two court rulings do not make every AI company's use of content legal under fair use. What makes these cases notable is that they are the first to issue substantive legal analyses on the issue; AI companies and publishers have been duking it out in court for years now. But just as Chhabria referenced and responded to the Anthropic ruling, all judges use past cases with similar situations as reference points. They don't have to come to the same conclusion, but the role of precedent is important. It's likely that we'll see these two rulings referenced in other AI and copyright/piracy cases. But we'll have to wait and see how big of an effect these rulings will play in future cases -- and whether it's the warnings or greenlights that hold the most weight in future decisions. For more, check out our guide to copyright and AI.
San Francisco Chronicle
an hour ago
- San Francisco Chronicle
Marc Benioff says AI does up to 50% of Salesforce's work and predicts ‘digital labor revolution'
Artificial intelligence now performs up to half the internal work at Salesforce, according to founder and CEO Marc Benioff, who says his company is spearheading a 'digital labor revolution' that will reshape the future of work. 'AI is doing 30% to 50% of the work at Salesforce now,' Benioff said during an episode of The Circuit with Emily Chang on Thursday, citing software engineering and customer service as key areas transformed by automation. He added that the company's internal use of AI has helped reduce hiring needs while boosting productivity. Benioff's bullish stance on AI is centered around a new platform called Agent Force, which deploys 'digital employees' to handle tasks ranging from customer service and analytics to marketing and branding. Salesforce aims to reach one billion active agents by the end of the year. 'It's the fastest-growing, most exciting thing we've ever done,' he said, noting that more than 5,000 customers are already using the technology. Benioff said the company's flagship AI agent has reached 93% accuracy in customer interactions, including with major clients like Walt Disney Co. Still, he emphasized the importance of vigilance, especially around security and misinformation. 'You have to be completely paranoid,' he said. At 59, Benioff continues to embrace his role as a tech industry disruptor. A self-taught programmer who sold his first software product as a teenager, he worked under Larry Ellison at Oracle before founding Salesforce in 1999. The idea for the company, he said, came to him while swimming with dolphins in Hawaii. Salesforce went on to revolutionize enterprise software by delivering it via the internet, pioneering the software-as-a-service model and becoming a global CRM leader with clients like Apple, Amazon and Boeing. Despite his enthusiasm, he acknowledged the technology's disruptive impact on labor. Salesforce has already cut over a thousand jobs, and Benioff predicts that today's CEOs may be the last to manage entirely human workforces. He estimates AI could unlock $3 to $12 trillion in global productivity. Benioff said he meditates daily, weaves Hawaiian spiritualism into company culture, and once spent over $20 million to license Einstein's likeness for Salesforce's AI branding. He proudly describes himself as 'the Taylor Swift of tech.' To illustrate his embrace of AI, Benioff said he co-authored the company's latest business plan with an AI assistant — an exercise he described as making the CEO job 'less lonely.' He is also blunt when it comes to competition. Benioff dismissed Microsoft's Copilot as 'repackaged ChatGPT,' drawing a contrast with Salesforce's more autonomous Agent Force platform. He supports breaking up Big Tech, saying the industry has 'probably been too big.' While facing criticisms over shifting diversity initiatives and political opportunism, Benioff remains committed to corporate responsibility. You can 'do well and do good,' he said, referring to Salesforce's 1-1-1 model, which pledges 1% of company equity, product and employee time to philanthropic causes. 'I think it's a false narrative,' he said. 'The people who are creating the future are here. This isn't the first gold rush we've been through.'
Business Wire
2 hours ago
- Business Wire
CB Insights Unveils Industry's First AI Agent Workforce for M&A, Partnerships, Competitive Intelligence, Deal Sourcing, GTM, and More
NEW YORK--(BUSINESS WIRE)--CB Insights, the leading provider of predictive market intelligence, today announced the launch of its Team of Agents – the industry's first comprehensive AI agent workforce designed specifically to enable business leaders with always-on market intelligence for strategic decision-making. CB Insights' growing team of agents starts with 11 specialized agents that work continuously to uncover insights, translate proprietary data into action, and accelerate smart decisions by providing the external intelligence layer that enterprises need to compete effectively across business development, investments, partnerships, GTM, and competitive intelligence. "The Scouting Report Agent is amazing – it's a huge time saver and lets us respond quickly with enough insight to get back to a client or be dangerous with." Share While generic AI assistants struggle with business context and lack reliable external data, CB Insights' Team of Agents delivers validated intelligence and data science on over 10 million companies and 1,500 tech markets through purpose-built specialists that understand the nuanced requirements of strategic business functions. Enterprise clients are already hiring CB Insights' Team of Agents to streamline their workflows and accelerate decision-making across research, diligence, briefings, and more – saving months of work. "The Scouting Report Agent is amazing – it's a huge time saver and lets us respond quickly with enough insight to get back to a client or be dangerous with," said Bethany Boutin, Business & Market Intelligence Lead: Commercial Solutions + Digital at Evernorth Health Services, a subsidiary of The Cigna Group. "Most AI agents today are generalists trying to do everything – but when it comes to strategic decisions, you need specialists who understand your business context and have access to reliable external intelligence," said Manlio Carrelli, CEO of CB Insights. "Our Team of Agents represents the evolution from generic AI assistants to purpose-built intelligence operatives. They work around the clock on your behalf, each one an expert in their domain, all powered by our double-verified Business Graph and data science on the world's companies and markets – insights and data that are not available on ChatGPT, the Internet, or in any other agent." Addressing the External Intelligence Gap While most enterprise strategic decisions require external market intelligence about competitors, markets, and potential partners, existing AI tools often lack the specialized context and validated data necessary for high-stakes business decisions. CB Insights' growing Team of Agents solves this, starting with 11 specialized agents that deliver: Acquisition Hunter: Identifies strategic acquisition targets and early-stage disruptors before competitors notice them Business Relationship Analyst: Maps customer, vendor, and partner relationships to reveal strategic insights ChatCBI: Serves as an AI-to-AI interface, allowing enterprises to integrate CB Insights intelligence directly into their existing AI workflows Commercial Due Diligence Accelerator: Reduces due diligence timelines from weeks to hours with automated expert interviews and risk assessments Competitive Sentinel: Monitors what competitors are buying, building, and investing in before it appears in financials Deep Analyst: Generates comprehensive research reports and market analyses in hours instead of weeks Earnings Analyst: Delivers CEO-ready competitive intelligence within hours of earnings calls Partner and New Business Finder: Detects high-value partnership and sales opportunities using funding and competitive signals Personal Briefing: Scours global news to surface personalized growth opportunities and competitive threats Sales Account Planner: Accelerates rep ramp-up with deep account understanding and tailored pitch strategies Scouting Report: Generates analyst-grade SWOT analyses on any private company on command The agents listed are not exhaustive and CB Insights' clients can create their own with Microsoft Copilot, MCP, and soon directly in ChatCBI. Agents Powered by the World's Most Comprehensive Business Graph Unlike AI agents that rely on public web data or limited datasets, CB Insights' Team of Agents operates on the proprietary Business Graph – a continuously validated dataset with patented data science covering management teams, funding rounds, partnerships, deals, earnings calls, hiring patterns, patents, and more across global business activity. "The quality of AI output depends entirely on the quality of data input," said Carrelli. "Our agents don't just scrape the web – they work with double-verified intelligence based on proprietary information collected from companies and their customers directly, unique datasets like their commercial relationships, and proprietary data science that scores future potential. This means when our Competitive Sentinel tells you a rival is planning a major acquisition, or when our Acquisition Hunter identifies a strategic target, you can make decisions with confidence." Designed for Enterprise Integration The Team of Agents operates across multiple platforms where strategic work happens: Native integration with CB Insights' Strategy Terminal for comprehensive market intelligence workflows Microsoft 365 Copilot integration for seamless incorporation into existing productivity tools Model Context Protocol (MCP) compatibility for any MCP-compatible AI agent to tap into CB Insights' datasets and tools – including ChatCBI – without a single line of code API and custom solutions for embedding intelligence into enterprise applications and workflows CB Insights' forward-deployed strategists work with you to automate your research and strategy flows with custom agents "We're not asking enterprises to change how they work – we're making their existing workflows dramatically more intelligent," said Carrelli. "Whether you're in Excel building a market model, in PowerPoint preparing for a board meeting, or doing deep work inside the CB Insights terminal, our agents provide the external intelligence you need exactly when and where you need it." Availability and Pricing CB Insights' Team of Agents is available immediately to enterprise customers through flexible deployment options. Individual agents can be deployed for specific use cases, or enterprises can access the full team for comprehensive market intelligence automation. For more information about CB Insights' Team of Agents and to request a demonstration, visit Headquartered in New York City, CB Insights is the leading provider of AI for market intelligence. The company aggregates, validates, and analyzes hard-to-find private and public company data. Its powerful AI tells users what it all means to them personally. The world's smartest companies rely on CB Insights to focus on the right markets, stay ahead of competitors, and identify the right targets for sales, partnership, or acquisition. Visit for more information.
