chatbots

A newly published research paper analyses how web and mobile conversational AI agents handle user privacy. The study, 'Prompt Like a Butterfly, Sting Like a Tracker', probes what data these chatbots and assistants collect and expose, and is drawing attention on technology discussion forums where readers are debating its findings on AI tracking practices.

The Trump administration has introduced an AI-powered 'concierge' on America.gov, and reporters who tested it describe what the chatbot can and cannot do. The tool is meant to help users navigate government information, and its launch has prompted discussion about how the federal government will use artificial intelligence in public-facing services.

Anthropic's Claude is being used to power coordinated teams of AI agents, with new structured setups letting multiple instances divide work, communicate and complete complex tasks together. The approach moves Claude beyond a single chatbot into an orchestration layer for multi-agent workflows. Commenters are debating how reliable such agent teams are in practice and what the shift means for automation of knowledge work.

A new research effort reported via Newswise focuses on enabling artificial intelligence systems to recognize the limits of their own knowledge, flagging when they cannot reliably answer a question rather than generating confident but wrong responses. The work addresses the well-known problem of AI hallucinations, a key trust barrier for chatbots and other AI tools used in science, medicine and everyday search.

SpaceX has introduced new pricing for Grok, the xAI chatbot, and Barron's reports on what the move could mean for the company's stock. The coverage ties the AI pricing decision to investor sentiment around Elon Musk's ventures and the valuations of SpaceX and xAI, which remain privately held.

A news feature explains 'red teaming', the practice of stress-testing AI systems, software and security defences by deliberately trying to break them. Practitioners probe chatbots and networks for flaws, biases and vulnerabilities before the public finds them, work often described as thankless but essential to keeping the internet safe.

Security commentary is warning that employees who use AI chatbots, assistants, or browser tools to speed up their work without their employer's knowledge—a practice dubbed 'shadow AI'—could be putting sensitive company data at risk. Malwarebytes has published an explainer on the trend, framing it as a growing cybersecurity concern for businesses.

A chatbot associated with Donald Trump is declining to answer when asked who won the 2020 US presidential election, according to Axios. The AI gives no response on the question rather than confirming Joe Biden's victory or repeating Trump's false claims of a stolen election. The avoidance is drawing attention as observers question how the tool was designed and what it reveals about how Trump's camp wants election questions handled.

Investor Kassan has backed Generative Engine Optimization startup OptimizeGEO in a deal valuing the company at $80 million. The funding highlights growing investor interest in GEO, a fast-emerging field focused on helping brands stay visible in AI-driven search results, as companies race to adapt to search shifting from traditional engines toward AI assistants and chatbots.

Users attempting to access Anthropic's Claude chatbot are being met with an "at capacity" message, suggesting demand has outstripped the service's available computing resources. The outage has prompted discussion among tech commentators about how quickly the AI assistant's popularity is growing and whether Anthropic can scale its infrastructure to keep up with sign-ups.