AI News: Tech Giants Face New York City Council Over Safety Rules

Nic Reeve14:00 05.10.20265 min read

Google, Meta, OpenAI and Anthropic representatives are appearing before New York City lawmakers on October 5, 2026, for a rare hearing focused on artificial-intelligence safety, accountability and possible local regulations. The proceeding is the latest development in the city’s push to question major technology companies directly, making it a major moment in AI news.

Why is the New York City Council holding the hearing?

The Council is examining whether fast-moving AI products create risks that existing safeguards do not adequately address. The October 5 hearing brings all 51 Council members together and is being conducted by the Committee of the Whole, allowing lawmakers to question companies in a single public session.

Event date: October 5, 2026.
Venue: New York City Council hearing in New York City.
Participants: Representatives of Anthropic, OpenAI, Google and Meta.
Format: Public testimony under oath.
Scope: AI safety risks, consumer protection and possible legislative safeguards.

According to the New York City Council, the hearing is intended to examine “the potential dangers of AI technology” and gather company input on legislative responses. Council Speaker Julie Menin announced the session on September 28, after earlier letters sought testimony from leading AI companies.

Which technology companies agreed to testify?

Anthropic, OpenAI, Google and Meta agreed to send representatives. Meta committed a senior executive before the other three companies confirmed participation after the Council warned that subpoenas could be used to compel testimony.

Anthropic: The company behind the Claude family of AI models.
OpenAI: The creator of ChatGPT and other generative-AI systems.
Google: The Alphabet company developing Gemini and related AI services.
Meta: The owner of Facebook, Instagram and WhatsApp, which is developing its own AI products.

According to Council announcements reported on September 28, OpenAI and Google agreed to appear on the Sunday before the hearing. Reporting published by CBS News on October 4 identified executives from the four companies as scheduled witnesses.

Why did subpoenas become part of the dispute?

The companies did not all accept the invitations at the same time. Speaker Menin initially requested participation from senior leaders, then warned that the Council could use its subpoena authority. OpenAI, Google and Anthropic later agreed to send representatives, while Meta had already committed to attend.

September 16: The Council announced plans for an October 5 hearing and requested participation from major AI leaders.
September 25: The Council unveiled proposed legislative measures and reiterated that attendance was expected.
September 28: The Council said OpenAI, Google, Anthropic and Meta would testify under oath.
October 5: The public hearing is scheduled to take place.

According to PoliticsNY, OpenAI, Google and Anthropic agreed to appear after lawmakers threatened subpoenas. The New York City Council separately announced that Elon Musk’s AI company, SpaceXAI, faced a subpoena after it did not confirm participation.

What will lawmakers ask the companies?

The Council has not published a complete question list, but its announcements identify several areas of concern. Lawmakers are expected to focus on how companies test models, respond to harmful outputs and protect New Yorkers from misuse.

How companies identify and reduce safety failures before releasing AI systems.
What protections exist for children and other vulnerable users.
How platforms respond to self-harm content, fraud, impersonation and manipulated media.
Whether companies provide enough transparency about model limits and incidents.
What city governments can regulate without conflicting with state or federal law.

The hearing follows reports of tens of thousands of AI-related safety incidents cited in recent coverage, although the public material gathered for the hearing does not establish a single official incident count. The Council’s stated purpose is broader: to assess risks and consider safeguards for residents.

What legislation is New York City considering?

The Council has linked the hearing to a package of proposals aimed at reducing potential harm from AI systems. The measures remain proposals, not enacted city law, and the public hearing is intended to inform debate over their contents and enforceability.

Rules addressing the safety and accountability of AI developers.
Possible requirements for companies to disclose risks or testing practices.
Consumer protections for people affected by automated decisions or generated content.
Local responses to harmful or deceptive uses of generative AI.

According to the Council’s September 25 announcement, the proposed bills were scheduled for discussion at the October 5 Committee of the Whole hearing. The Council has also questioned whether protections at the state and federal levels are sufficient for New Yorkers.

What makes the hearing unusual?

The proceeding combines an unusually large group of lawmakers with testimony under oath from several leading AI companies. According to the Council, it is the first time major AI firms have been set to provide public testimony under oath before the body following recent incident reports.

51 Council members: The full City Council is expected to participate.
Four companies: Anthropic, OpenAI, Google and Meta have confirmed attendance.
Under oath: The format gives testimony a more formal legal setting than a private briefing.
Local focus: Questions will center on risks experienced by people in New York City.

Fortune reported on September 25 that the hearing would be a Committee of the Whole, a format the Council had not used since 2022. The same report said several requested chief executives were unlikely to appear personally, meaning company representatives may answer questions instead.

What happens after the testimony?

The Council can use the testimony to revise proposed bills, request further records or pursue additional hearings. The hearing itself does not automatically create new rules. Any measure would still need to move through the Council’s legislative process and comply with higher-level law.

Lawmakers may compare company safety policies and public commitments.
The Council may seek documents or additional testimony.
Proposed legislation could be amended after the hearing.
Companies may face continuing scrutiny over AI-related incidents in New York.
The subpoena dispute could shape future dealings between City Hall and technology firms.

The hearing also places pressure on companies to explain their safeguards in a public forum. Their answers may influence how New York City approaches AI oversight while broader debates continue in Albany and Washington.

Sources

AI & Tech

AInews: Majority of Americans Now Say Artificial Romance Counts as Cheating

On 3 August 2026, a new survey titled The State of AI Romance 2026 reported that 50.5% of U.S. adults consider a partner’s romantic or sexual relationship with artificial intelligence to be cheating, giving AInews a fresh data point in the debate over digital fidelity. What exactly did the new survey on AI romance and cheating find? The August 2026 survey by AI Girlfriend Coach asked U.S. adults whether they would see a partner’s romantic or sexual relationship with an AI bot as infidelity. Just over half said yes, while many of the rest still felt uneasy even if they stopped short of calling it cheating. According to AI Girlfriend Coach’s The State of AI Romance 2026 study, fielded in August 2026 among U.S. adults: 50.5% of respondents said a partner’s romantic or sexual relationship with an AI would count as cheating. 21.2% said it would not be cheating but that they would be uncomfortable with it. Combined, 71.7% expressed discomfort with a partner having an AI-mediated romance. The study examined how adults view AI girlfriends, AI boyfriends and emotional attachment to digital companions, concluding that artificial partners are already reshaping ideas of loyalty and betrayal in long-term relationships. How do these findings compare with other recent research on AI companions and infidelity? Other surveys from 2025 and 2026 show a similar pattern: large numbers of Americans treat romantic or sexual interaction with AI as crossing a line, although they differ in whether they label it outright cheating or a lesser breach of trust. Recent national and industry studies report the following key figures: According to a 2,000-person U.S. singles survey conducted by OnePoll for dating app Plenty of Fish in 2025, 54% of respondents said romantic or emotional engagement with AI chatbots constitutes infidelity. The same Plenty of Fish study found that this 54% is just one percentage point below the 55% who classify emotional bonds with another human as cheating, putting digital and human emotional affairs on nearly equal footing. A study by the Kinsey Institute and DatingAdvice.com, fielded with 2,000 U.S. singles in 2025, reported that 61% said sexting or falling in love with an AI “crosses the line” in a relationship. According to a breakdown of that Kinsey–DatingAdvice survey published in 2026, 32% called sexting with an AI chatbot cheating, while 29% said an AI romance itself counts as cheating. A July 2026 article discussing Talker Research results said 24% of Americans consider flirting with AI a form of cheating, with the figure rising to 36% among Generation Z respondents. A July 2026 analysis on romantic AI companions reported that roughly one third of Americans say a partner’s use of such a companion would count as cheating. Taken together, the new 50.5% figure places AI romance near the top of the range of existing estimates, aligning with earlier research that shows between one quarter and just over half of Americans equate AI intimacy with betrayal. Who is most likely to see AI relationships as cheating, and who is more accepting? Age and gender sharply divide perceptions of AI romance. Younger adults and women tend to view AI partners as a breach of fidelity. Older adults and some young men are more open to parallel digital relationships, or see them as a way to prevent traditional cheating. Different studies highlight these demographic splits: A 2025 national study of Americans aged 18–29 found that 56% in this group say having a romantic or intimate relationship with AI is cheating. In a 2025 survey about AI boyfriends and girlfriends, 25.34% of women regarded AI dating as outright cheating, compared with 18.53% of men. The same AI partner survey reported that 30% of millennials see AI partners as cheating, compared with 15% of Gen Z respondents. That study also noted that 17% of Gen Z males believe AI partners can help prevent cheating in human relationships, suggesting some see digital romance as a safety valve rather than a betrayal. A 2025 study on “Artificial Romance” found that adults aged 60 or older were the most accepting of AI romantic relationships alongside human relationships, with 50.29% saying these parallel relationships are not cheating on human partners. According to that same research, younger adults aged 18–29 were the most likely to judge simultaneous AI and human romantic relationships as cheating and unacceptable. The August 2026 AI Girlfriend Coach survey did not publish a full demographic breakdown in its press summary, but its headline figure aligns with earlier findings that a majority of Americans either condemn AI romance outright or feel deeply uneasy about it. Are people actually forming secret relationships with AI chatbots? Surveys about behaviour show that AI affairs are not merely hypothetical. Young adults in committed relationships increasingly report having hidden romantic attachments to AI chatbots, often without their partners’ knowledge. Research from Brigham Young University’s Wheatley Institute, published in May 2026, examined AI attachments among 18- to 30-year-old Americans who were dating, engaged or married. According to the BYU Wheatley Institute study, one in seven partnered young adults has a secret and frequent romantic attachment to an AI chatbot. A detailed summary of the same research reported that the survey covered 2,431 American adults between 18 and 30 in romantic relationships. Separate polling on AI companions found that at least one in six American adults has interacted at least once with a romantic AI companion, while about one third said a partner’s use of such an AI would count as cheating. These behavioural data points show that the cheating dilemma is no longer theoretical. Digital partners sit inside phones and laptops across the United States, sometimes hidden from human partners who may still assume betrayal only happens with other people. How do Americans define cheating when technology blurs the lines? Researchers say Americans now draw distinctions between types of AI interaction, splitting flirting, sexting and full-blown romantic attachment into different categories of moral violation. Emotional entanglement with AI appears to worry many respondents as much as physical infidelity. Recent surveys illustrate how definitions are evolving: Plenty of Fish’s 2025 survey reported that 54% of U.S. singles classify romantic or emotional engagement with AI chatbots as infidelity, almost matching the 55% who say emotional bonds with another human count as cheating. A 2025 study covered by Mashable found that 33% of 2,000 American adults believe that sexting or entering a romantic relationship with AI constitutes cheating. In that Mashable-reported survey, 64% of those who saw sexting or AI romance as cheating said both behaviours are infidelity, 21% believed only sexting with AI is cheating, and 15% said only a romantic relationship with AI counts. The Kinsey Institute and DatingAdvice.com’s 2025 national study found that 61% of singles see sexting or falling in love with an AI as cheating, underscoring broad moral concern around digital intimacy. A detailed reanalysis of those Kinsey–DatingAdvice data in 2026 highlighted that, when asked separately, 32% called sexting with an AI chatbot cheating, and 29% called an AI romance cheating. Talker Research’s July 2026 survey, reported in Italian media, said 24% of Americans consider flirting with AI cheating, with the share rising to 36% among Gen Z. These figures suggest that while outright physical betrayal with another human still dominates common-sense definitions of cheating, a growing share of Americans view emotional and sexual interaction with AI as falling inside the same moral category. What does this mean for couples negotiating boundaries around AI in daily life? Therapists and relationship researchers argue that couples now need to talk explicitly about AI companions, just as they might discuss social media behaviour or contact with ex-partners. Silence leaves room for mismatched expectations and sudden feelings of betrayal. Recent commentary and clinical research highlight several practical implications for relationships: According to the AI Girlfriend Coach report, the reasons people turn to AI companions appear to be “less about sex and romance than commonly assumed”, suggesting some are seeking emotional support or nonjudgmental conversation. Counsellors writing in a 2025 study on artificial romance note that older adults might see AI partners as companionship that does not threaten their human relationships, which could clash with younger partners’ stricter expectations of fidelity. Deseret News coverage of the BYU Wheatley Institute study described real-world couples confronting the discovery of hidden AI relationships, raising questions about whether confession and forgiveness work differently when no human third party is involved. A July 2026 analysis on flirting with AI companions emphasised that at least one in six adults has tried romantic interaction with AI, meaning many couples are negotiating this issue for the first time without established social norms to guide them. Radio and podcast discussions of the October 2025 cheating survey warned that younger Americans, especially those aged 18–29, are already more likely than older generations to classify AI romance as cheating, hinting at intergenerational friction inside families. The new 50.5% figure adds weight to calls from therapists and relationship coaches who say partners should now address AI boundaries early, and revisit them as technology evolves, rather than waiting for secret digital attachments to surface during conflict.

Nic Reeve·14:00 04.09.2026

AI & Tech

Anthropic Gives Claude Cowork Shared Memory with Chat for Persistent Context

Anthropic is rolling out a major upgrade to its AI assistant, giving Claude Cowork the ability to seamlessly reuse information it learns in regular chat. The company has merged the memory systems behind Claude’s chat interface and its Cowork desktop agent, so details you share in one surface can now automatically be used in the other. One Shared Memory Across Chat and Cowork Previously, Claude’s long‑term memory was largely confined to chat sessions and was synthesized periodically, meaning it could take up to a day before information carried over into new conversations. Cowork, which runs complex, multistep jobs on a user’s desktop or in the cloud, relied on its own background memory file and prompt stitching to simulate continuity. With the August 25 update, Anthropic has combined these mechanisms into a single, shared memory system that serves both chat and Cowork. Anthropic describes the change simply: the same memory now powers both Claude chat and Cowork. When users hand a task to Cowork—such as drafting reports, updating spreadsheets, or coordinating project documents—the context Claude has accumulated over months of chats is immediately available. Likewise, any new facts or preferences learned during Cowork runs are written back into the shared memory and become available in subsequent chat sessions. Real‑Time Memory, Not Just End‑of‑Chat Summaries Another important shift is how Claude updates memory. Instead of waiting to summarize an entire conversation once it ends, Claude now adds topics to memory in real time as users chat. This means that if a user mentions that a project deadline moved to September, that update can be reflected in memory almost immediately and show up in the very next interaction—whether in chat or Cowork—without requiring a manual “remember this” command. Anthropic’s support materials explain that when Cowork runs in the cloud, what Claude remembers from previous chats is automatically available, and what emerges during Cowork tasks feeds back into chat memory. Behind the scenes, each Cowork prompt is assembled from the user’s immediate request, their global instructions, and a relevant slice of the shared memory, allowing the AI to behave as if it has persistent awareness of roles, projects, and preferences. What Users Gain: Less Repetition, More Continuity The practical effect for users is that they no longer need to repeatedly brief Claude on who they are, what they are working on, or how they like to work every time they switch between chat and Cowork. Anthropic and independent commentators highlight several common scenarios: Persistent project context: Ongoing details such as quarterly goals, client names, and current project status can be retained across weeks or months and recalled in both chat and Cowork. Stable roles and preferences: If a user identifies themselves as an investment analyst, a teacher, or a particular type of creator, Claude can remember that role and tailor responses accordingly, even when individual chats are short or focused on different tasks. Cross‑device consistency: The shared memory applies across web, desktop, and mobile experiences, so moving from a browser chat to the Cowork desktop agent no longer breaks context. Tech industry observers note that this update positions Claude more directly as an AI “teammate” that can track medium‑ and long‑term workstreams instead of acting purely as a session‑bound chatbot. Transparency and User Control Over Memory The shared memory system arrives alongside a push for greater user control. Anthropic now surfaces everything Claude remembers in a dedicated Topics view within memory settings, where users can inspect, edit, or delete individual entries. Memory is stored as discrete, categorized entries rather than a single opaque summary, making it easier to remove outdated or inaccurate information. Users can also pause memory or reset it entirely if they no longer wish Claude to retain prior context. In addition, Anthropic provides guidance on importing and exporting memory, so the information Claude stores about a user is not locked in and can in principle be backed up or moved. Handling Sensitive Topics Anthropic has emphasized that the system is designed to minimize the capture of highly sensitive information by default. Topics such as health data, beliefs, and other potentially sensitive categories are excluded from memory unless users explicitly opt in via an “Include sensitive topics in memory” setting. For business customers, team or enterprise administrators can centrally control whether memory is enabled at all, and may choose more restrictive policies depending on corporate governance requirements. External reporting indicates that memory generation is turned on by default for free, Pro, and Max plans, while Cowork itself is not available on free accounts. For organizations that want to keep different workstreams separated, Anthropic has indicated that the only way to maintain fully separate memories for chat and Cowork is to use different accounts, since the new system treats them as a single unified space. Availability and Limitations The new shared memory capability began rolling out on August 25, 2026, across Claude’s web, desktop, and mobile experiences, as well as Cowork running in the cloud. Earlier in the year, memory support was limited to chat surfaces, and some third‑party analyses noted that Cowork lacked access to that long‑term context. Anthropic’s latest release notes and help center now explicitly state that memory works across both chat and Cowork when the latter runs in the cloud environment. There are still technical constraints. Cowork’s use of memory depends on cloud execution rather than purely local processing, and incognito or memory‑disabled sessions remain stateless by design. As with other AI systems, Anthropic cautions that Claude’s memory is selective: it prioritizes high‑level preferences and recurring topics rather than storing every detail of every conversation. A Step Toward More Personalized AI Workflows By unifying memory between Claude chat and Cowork, Anthropic is betting that users will value a more personalized and continuous AI experience, particularly for complex, ongoing work. The update reduces friction for individuals juggling multiple projects and gives enterprises a clearer path to building AI‑augmented workflows that persist over time. At the same time, the company is attempting to balance convenience with privacy and security by giving users fine‑grained controls and limiting sensitive data retention by default.

Nic Reeve·16:00 26.08.2026

AI & Tech

OpenAI Says AI’s Next Economic Test Is Turning Big Ideas Into Finished Work

OpenAI published “The eternal complement” on October 1, 2026 , an essay arguing that advanced artificial intelligence may create its greatest economic value through the routine work required to turn ambitious ideas into functioning systems. The discussion adds a new angle to AI news by focusing less on machine-generated breakthroughs and more on execution, coordination and institutional capacity. What did OpenAI publish? OpenAI released the essay through its Intelligence Age platform on October 1, 2026. The article was written by Hemanth Asirvatham and Elliott Mokski, who present a framework for understanding how frontier intelligence interacts with the organizations and systems needed to apply it. The publication is an essay, not a product announcement, model launch or performance benchmark. OpenAI’s newsroom lists “The eternal complement” as part of its Intelligence Age material, while Unite.AI describes it as the first entry in a planned series about the next economy. Publication date: October 1, 2026, according to OpenAI’s Intelligence Age index. Authors: Hemanth Asirvatham and Elliott Mokski, according to Unite.AI. Format: An analytical essay rather than a software or hardware release, according to OpenAI’s newsroom listing and reporting by Remio.ai. What does “eternal complement” mean? The essay uses the economic idea of complements, or inputs whose value rises when they are used together. Asirvatham and Mokski apply that concept to advanced machine intelligence and the practical ability to carry ideas through development, regulation, funding and production. According to Unite.AI’s October 1 report, the authors call this often-overlooked execution capacity “institutional intelligence.” The term covers the laws, bureaucratic processes, financing arrangements, supply chains and specialist coordination that stand between a promising concept and a usable result. The central argument is direct: smarter systems do not remove the need for execution. They can increase the value of people and institutions capable of carrying out difficult, repetitive or highly coordinated work. Why does the essay focus on routine work? The publication argues that advanced systems may matter most when they reduce the cost of work that is necessary but rarely celebrated. The essay points to activities such as writing code, searching unfamiliar research and converting rough concepts into working prototypes, according to Unite.AI. Remio.ai’s October 2 analysis characterizes the argument as a shift from asking whether AI can generate an answer to asking whether the answer can become a completed project. That process can require repeated testing, documentation, approvals, hiring, procurement and technical integration. Routine does not mean unimportant. A discovery that cannot be tested, financed or deployed has limited practical effect. Research: Systems can help locate and interpret unfamiliar literature, according to Unite.AI. Software: AI can assist with code and early prototypes, according to Unite.AI. Implementation: Organizations still need to manage rules, resources, specialists and supply chains, according to the essay’s description in Unite.AI and Remio.ai. Is OpenAI announcing a new product? No. The October 1 publication does not announce a model, subscription, API, benchmark or commercial deployment. OpenAI’s newsroom places the piece in the Intelligence Age section, and Remio.ai explicitly describes it as a framework for judging how advanced systems might translate ideas into broader material progress. That distinction matters for readers following the company’s product releases. The essay sets out a view about economics and organizational change. It does not provide a new capability that customers can download or test. OpenAI’s presentation also says the piece reflects the authors’ perspective rather than serving as a formal product specification, according to MediaRelease.co’s report on the publication. What could change for workers and organizations? The argument places execution-heavy occupations at the center of the next phase of automation. Workers who translate goals into code, reports, prototypes, compliance documents or operating procedures could gain tools that shorten the path from instruction to completed task. That does not mean every role disappears. The framework instead suggests that the balance of work may change. People could spend less time producing first drafts and more time setting objectives, checking outputs, making judgments and handling exceptions. Organizations face a parallel challenge. Faster generation can expose bottlenecks in approval systems, budgets, data access and procurement. A company may possess powerful software and still move slowly if its internal processes cannot absorb the output. Potential beneficiaries: Small teams that can use machine assistance to perform work once requiring larger departments, according to Unite.AI. Potential pressure points: Compliance, management, infrastructure and supply-chain functions that determine whether an idea reaches deployment, according to the essay’s framework. Human responsibility: Goal-setting, verification and decisions involving risk remain central to the model described by the authors. How does the publication fit OpenAI’s wider agenda? OpenAI launched its Intelligence Age platform in August 2026, according to Unite.AI, presenting it as a venue for perspectives on advanced AI and the next economy. “The eternal complement” is listed as the platform’s first essay in the series. The timing places the article within a wider debate about what increasingly capable systems will change beyond chat interfaces and software demonstrations. Rather than measuring progress only through benchmark scores, the essay asks whether institutions can convert technical ability into discoveries, products and public services. That question extends beyond OpenAI. Governments, universities, companies and investors all control parts of the execution chain described in the essay. Their rules and capacity will influence how quickly machine-assisted ideas become real-world outcomes. What happens next? The immediate next step is further discussion through the Intelligence Age series. No launch date, business target or measurable forecast accompanied the October 1 essay, so its claims remain a strategic argument rather than a tested prediction. Readers will likely judge the thesis against practical evidence: whether smaller teams can complete more complex projects, whether approval systems adapt to faster technical work and whether organizations invest in the less visible infrastructure needed for deployment. The essay’s test is practical. Can intelligence survive contact with institutions?

Nic Reeve·14:00 03.10.2026

AI News: Tech Giants Face New York City Council Over Safety Rules

Why is the New York City Council holding the hearing?

Which technology companies agreed to testify?

Why did subpoenas become part of the dispute?

What will lawmakers ask the companies?

What legislation is New York City considering?

What makes the hearing unusual?

What happens after the testimony?

Sources

Related Articles

AInews: Majority of Americans Now Say Artificial Romance Counts as Cheating

Anthropic Gives Claude Cowork Shared Memory with Chat for Persistent Context

OpenAI Says AI’s Next Economic Test Is Turning Big Ideas Into Finished Work