U.S. House of Representatives
Oct 29, 2026

House Democrats ask Trump to preserve AI chatbot conversations

Techub News, citing Crypto Briefing, reported that Democratic lawmakers in the U.S. House of Representatives have asked Donald Trump to preserve records of his conversations with AI chatbots. The request points to growing concern over the expanding role of artificial intelligence in governance. It also adds to calls for tighter oversight and stronger transparency standards around digital communications. The report did not provide additional details on the scope of the request or any timeline tied to it.

450
House Democrats ask Trump to preserve AI chatbot conversations
Mark Cuban
Oct 19, 2026

Mark Cuban says AI memory can lock users into a single chatbot

Billionaire investor Mark Cuban said AI memory features can make it harder for consumers to switch between chatbots, arguing that user lock-in will remain high until a standardized format exists for exporting model memory. In a post on X, Cuban said the more often a person uses a model, the more data it stores, and the harder that data becomes to move elsewhere. He did not say migration is impossible, but argued that for ordinary users, transferring AI memory is far more complicated than saving an Excel or Word file. The report also pointed to recent moves by Anthropic. In March this year, the company opened Claude’s memory feature to free users and allowed new users to paste chat histories from rivals such as ChatGPT. It also noted that Anthropic’s open standard, Model Context Protocol, or MCP, introduced in 2024, could in theory allow AI memory to exist outside any single chatbot.

890
Mark Cuban says AI memory can lock users into a single chatbot
AI
Oct 10, 2026

Physicists Propose a Formula to Estimate When an AI Chatbot May Turn Harmful

Physicists at George Washington University say they have developed a formula that estimates how many safe answers an AI chatbot will produce before it slips into a harmful one. The study, by Neil Johnson and Frank Yingjie Huo and published in Patterns, builds on a preprint first posted in February. Their approach ties the shift to the model’s attention head, the mechanism that weighs earlier words when generating the next token. In the preprint, the method correctly identified whether a model would fail immediately or only after a run of acceptable output in 15 of 16 clear-cut tests, or 94%. The work focuses on open-weight models from OpenAI, EleutherAI, and Meta, with the preprint testing systems ranging from 124 million to 410 million parameters. The published paper reportedly extends that to seven models up to 12 billion parameters. The researchers are especially interested in on-device AI that runs fully offline on phones or laptops, where no cloud-based moderation layer checks model output. They propose a low-cost parallel monitor that warns when a safety threshold is approached and also describe ways to push the tipping point farther away, while arguing that alignment training alone cannot remove the underlying mechanism.

1060
Physicists Propose a Formula to Estimate When an AI Chatbot May Turn Harmful
Circuit Breaker Labs
Oct 2, 2026

Circuit Breaker Labs builds safety testing model for mental health chatbots

AI safety testing startup Circuit Breaker Labs said it is building a "crash test dummy" model for mental health chatbots, according to a Techub News item citing Crypto Briefing. The model is designed to simulate user interactions so the company can test and improve chatbot reliability. The stated goal is to reduce potential psychological harm to vulnerable users. The announcement centers on safety evaluation for chatbot systems used in mental health settings, where reliability and response quality can carry higher stakes for users.

860
Circuit Breaker Labs builds safety testing model for mental health chatbots
China regulation
Oct 2, 2026

Chinese regulators impose new guardrails on AI companion chatbots

Chinese regulators have introduced new guardrails for AI companion chatbots, according to Techub News, which cited Crypto Briefing. The move was described as a response to concerns over risks to social stability. The report said the new restrictions could reshape global technology standards and affect how AI products are developed, as well as how users engage with them. No additional details on the measures were disclosed in the source brief.

830
Chinese regulators impose new guardrails on AI companion chatbots
AI
Aug 16, 2026

California Bill Moves Ahead to Restrict AI Chatbots in Mental Health Care

California lawmakers have advanced Senate Bill 903, a measure that would restrict how artificial intelligence can be used in psychotherapy and mental health decision-making. The bill cleared the Assembly Appropriations Committee on Thursday in a 13-0 vote and now heads to a third reading, though no Assembly floor vote has been scheduled. Introduced on January 21, the proposal would bar AI chatbots from presenting themselves as therapists, limit AI use in psychotherapy to support functions, and require clinician oversight as well as patient consent. It also includes protections for therapy data, while carving out exemptions for peer support, religious counseling, research, and federally approved medical tools. The California Senate previously passed the bill 39-0 in May, and two Assembly policy committees later approved it by 17-0 and 14-1. Critics, including TechNet executive director Robert Boykin, argue that requiring clinician involvement could slow access to care at a time when every county in California faces a shortage of behavioral healthcare workers. The legislative push arrives as concern grows over AI’s role in mental health, with a June American Psychological Association survey finding that 77% of psychologists had patients using AI for mental health purposes.

1110
California Bill Moves Ahead to Restrict AI Chatbots in Mental Health Care
Apple
Aug 4, 2026

Apple unveils upgraded Siri AI at WWDC26 and opens developer beta

Apple introduced a Siri AI upgrade at WWDC26 on June 8, rolling out a new conversational version of the assistant and stepping into the chatbot race against ChatGPT. The new Siri uses a privacy-first architecture, and a developer beta is now available. The update marks Apple’s first ground-up rebuild of Siri since 2011, giving the company a major distribution edge through the assistant’s presence on hundreds of millions of preinstalled devices. According to CryptoBriefing, the release does not include any crypto technology or digital assets. It also does not support on-chain payments or Web3 wallets.

2340
Apple unveils upgraded Siri AI at WWDC26 and opens developer beta
European Union
Jul 31, 2026

EU to enforce parts of AI Act from Aug. 2, with new transparency rules taking effect

The European Commission said it will begin enforcing relevant provisions of the Artificial Intelligence Act from Aug. 2, working through the EU AI Office together with national authorities in member states. At the same time, new transparency obligations will come into force for certain AI systems. Those rules require users to be clearly informed when they are interacting with artificial intelligence and whether content has been generated or modified by AI. Interactive systems such as chatbots must identify themselves as AI rather than human operators. Deepfake images, video and audio created or edited with AI must also be labeled, and AI-generated or AI-modified material will need machine-readable markers so it can be identified and traced. The Commission said the measures are designed to reduce deception and manipulation, help the public make more informed judgments, and give companies clearer compliance requirements and implementation paths. On the same day, it also released a list of more than 180 organizations that have signed the first version of the Code of Practice on Transparency for AI-Generated Content, which is intended to support implementation of the AI Act’s transparency provisions.

2550
EU to enforce parts of AI Act from Aug. 2, with new transparency rules taking effect
Total 9 items
Go topage