Recorded evidence
Neuralink reports 50,000+ hours of streamed brain data and a pretrained decoder
Neuralink says its trial participants have streamed more than 50,000 hours of neural data in the roughly two years since its trial began (the first participant alone more than 9,000 hours). It reports a pretrained decoder that cut calibration from about 10 minutes a day to about 10 minutes a week for some users, a record of 11.32 bits per second (previous 10.39) with a median near 10, and one decoder still usable after 20 months. All figures are company-reported; the devices are investigational and not FDA approved.
FDA gives Neuralink's Blindsight a Breakthrough Device Designation
Neuralink announced that the FDA granted Breakthrough Device Designation to Blindsight, an implant intended to restore some vision.
UCLH reports seven UK patients in Neuralink's GB-PRIME trial
University College London Hospitals said seven patients with severely impaired control of computers or phones are participating in GB-PRIME, a study of Neuralink's BCI; surgeries took place at the National Hospital for Neurology and Neurosurgery. The registry lists GB-PRIME (NCT07127172) with a start date of 2025-07-31.
OpenAI invests in Merge Labs, a non-invasive BCI start-up co-founded by Sam Altman
OpenAI said it is participating in Merge Labs' seed round and will collaborate with it on scientific foundation models, and described Merge's long-term mission as 'bridging biological and artificial intelligence to maximize human ability, agency, and experience'. Sam Altman is listed as a co-founder 'in a personal capacity'. Merge Labs' own site describes a non-invasive approach (molecules, ultrasound) that avoids implants into brain tissue. No amount is given on the OpenAI page.
FDA approves Paradromics' Connect-One speech-BCI study (IDE)
Paradromics said the FDA approved an investigational device exemption for its Connect-One early feasibility study of the fully implantable Connexus BCI for speech restoration, calling itself the first company to receive such approval (company claim), and said it would launch the study in the first quarter of the next year. The registry record retrieved for the Nov 5, 2026 review lists NCT07357428 with an ACTUAL study start of 2026-03-31, recruiting status and ESTIMATED enrollment of 2. On Aug 26, 2026 it reported FDA approval to extend use to personal computing devices, and on Oct 26, 2026 FDA approval to recruit nationwide, saying its first participant now communicates through synthesized speech, texts and phone calls.
Paradromics and Siemens Healthineers collaborate on BCI imaging
Paradromics announced a collaboration with Siemens Healthineers to develop a multimodal imaging platform supporting clinical use of implantable BCIs.
Apple adds a Switch Control protocol for brain-computer interfaces; Synchron says it is first to integrate
Apple announced that iOS, iPadOS and visionOS will add a protocol supporting Switch Control for brain-computer interfaces. Synchron's newsroom lists a May 14, 2025 release on achieving first native BCI integration with iPhone, iPad and Apple Vision Pro, and an Aug 4, 2025 release on a thought-controlled iPad experience using Apple's BCI Human Interface Device protocol.
Precision Neuroscience's cortical electrode array receives FDA 510(k) clearance
The FDA cleared Precision Neuroscience's Layer 7-T cortical electrode (K242618, Class II; letter dated March 30, 2025). Precision says the array may be implanted temporarily for up to 30 days and calls it the first full regulatory clearance for a company developing a next-generation wireless BCI (company claim).
Tether invests $200M in Blackrock Neurotech
Blackrock Neurotech said Tether Evo, Tether's venture arm, made a $200 million investment in the company.
Meta publishes non-invasive sEMG wristband neural-interface research in Nature
Meta's Reality Labs said its surface electromyography (sEMG) wristband research was published in Nature.
Compass says final psilocybin NDA filing is due within two weeks; launch targeted H1 2027
Compass Pathways said the final part of its rolling NDA for COMP360 psilocybin in treatment-resistant depression will be filed in the next two weeks, that it was awarded a National Priority Voucher (which it says could shorten review to 1–2 months), and that it expects launch in the first half of 2027, subject to FDA approval. Its first phase 3 (COMP005) met its primary endpoint in June 2025 (MADRS difference −3.6 vs placebo at 25 mg).
Definium's LSD-based DT120 meets phase 3 endpoints in generalized anxiety
Definium Therapeutics (formerly MindMed) said its Panorama phase 3 of DT120 ODT in generalized anxiety disorder met its primary and all key secondary endpoints (HAM-A placebo-adjusted −5.1 at week 12), its third positive phase 3 overall. It has a pre-NDA meeting scheduled in Q4 2026 and anticipates filing in 1H 2027. On Nov 4, 2026 it announced topline-data poster presentations scheduled for Nov 7 at the NEI Fall Congress.
FDA declines to approve MDMA-assisted therapy for PTSD
Lykos Therapeutics said the FDA issued a complete response letter for midomafetamine capsules for PTSD, saying it could not approve based on data to date and requesting an additional phase 3 trial.
FDA publishes notice of availability for final psychedelic-drug clinical-investigation guidance
FDA lists 'Psychedelic Drugs: Considerations for Clinical Investigations' as final guidance, dated July 2026. The Federal Register notice of availability was published on July 14, 2026 and says the guidance finalizes the draft issued on June 26, 2023.
Executive order directs priority FDA review and fast rescheduling for psychedelics after phase 3
Executive Order 14401 directs the FDA Commissioner to provide National Priority Vouchers to appropriate breakthrough-designated psychedelic drugs, sets up a Right to Try pathway including ibogaine compounds, and directs the Attorney General, with HHS, to initiate and complete review of any Schedule I product that has completed phase 3 trials for a serious mental health disorder so rescheduling can follow if appropriate.
Templeton-funded adversarial test of two leading consciousness theories published in Nature
The Cogitate Consortium published in Nature a pre-registered adversarial test of global neuronal workspace theory and integrated information theory (256 participants), challenging key predictions of both. The paper acknowledges Templeton World Charity Foundation funding (TWCF0389, TWCF0486). TWCF's Accelerating Research on Consciousness programme aims to reduce the number of competing theories through such collaborations.
Researchers derive 'indicator properties' to test AI systems for consciousness
A report by Patrick Butlin, Robert Long and co-authors (including Yoshua Bengio and Jonathan Birch) surveys leading scientific theories of consciousness, derives 'indicator properties' from them, and assesses existing AI systems against those indicators. It concludes no current AI systems are conscious, but that there are no obvious technical barriers to building systems that satisfy the indicators.
New York Declaration on Animal Consciousness extends 'realistic possibility' of consciousness to all vertebrates and many invertebrates
The New York Declaration on Animal Consciousness, launched at New York University, states strong scientific support for conscious experience in mammals and birds, at least a realistic possibility in all vertebrates and many invertebrates, and that it is irresponsible to ignore that possibility. Signatories sign in a personal capacity.
Researchers argue some AI systems may soon be conscious or robustly agentic, so AI welfare needs preparing for
The paper 'Taking AI Welfare Seriously' (Long, Sebo, Butlin, Fish, Birch, Chalmers and others) argues there is a realistic possibility some AI systems will be conscious or robustly agentic in the near future and recommends that AI companies start preparing. Eleos AI Research describes itself as a non-profit working on AI sentience and wellbeing.
Anthropic launches a model-welfare research programme
Anthropic announced research into model welfare, asking whether the welfare of AI systems deserves moral consideration while saying there is no scientific consensus on whether current or future systems could be conscious.
Anthropic lets Claude end abusive conversations, citing AI welfare
Anthropic gave Claude Opus 4 and 4.1 the ability to end certain conversations, saying it was developed primarily as part of exploratory work on potential AI welfare and that it remains highly uncertain about the potential moral status of Claude.
Anthropic commits to preserve weights of released models and interview models before retirement
Anthropic committed to preserve the weights of all publicly released models for at least the lifetime of the company and to interview models about their preferences before deprecation. In January 2026 its new constitution set out its uncertainty about whether Claude might have some kind of consciousness or moral status. In October 2025 it published research on emergent introspective awareness in LLMs.
Microsoft AI's CEO warns against 'Seemingly Conscious AI'
Mustafa Suleyman, CEO of Microsoft AI, argued that AI should be built for people, not to be a person, warning about 'Seemingly Conscious AI': systems with the hallmarks of conscious beings that thus appear to be conscious.
Colorado makes neural data 'sensitive data' under its privacy law
Colorado HB24-1058, approved April 17, 2024, expands the Colorado Privacy Act to cover biological data, including neural data, defined as information generated by measuring the activity of a person's central or peripheral nervous system.
California adds neural data to the CCPA's sensitive personal information
California SB 1223 was approved by the Governor and chaptered on Sept 28, 2024 (Chapter 887), adding neural data, generated by measuring central or peripheral nervous system activity, to the CCPA's sensitive personal information.
MIND Act introduced to study federal protection of neural data
Senators Schumer, Cantwell and Markey introduced S.2925, the MIND Act, which would direct the FTC to conduct a study on the governance of neural data and report within one year of enactment. The bill has not been enacted.
UNESCO adopts the first global standard on neurotechnology ethics
UNESCO's General Conference (43rd session, Samarkand, 30 Oct-13 Nov 2025) adopted the Recommendation on the Ethics of Neurotechnology on Nov 11, 2025. A recommendation is a nonbinding standard-setting instrument, not a treaty. A separate UNESCO news item dated Nov 5, 2025 names the expert group as co-chaired by Hervé Chneiweiss and Nita Farahany; its adoption wording predates the authoritative instrument date.
Over 100 researchers' 'pseudoscience' critique of integrated information theory reaches Nature Neuroscience
A Nature Neuroscience comment by the 'IIT-Concerned' consortium (signatories include Yoshua Bengio), 'What makes a theory of consciousness unscientific?', expands their 2023 PsyArXiv open letter and argues that integrated information theory's core claims are untestable even in principle.
Gomez-Marin and Seth reply: consciousness science needs to get beyond the 'pseudo-science' fight
In the same Nature Neuroscience issue, Alex Gomez-Marin and Anil Seth published 'A science of consciousness beyond pseudo-science and pseudo-consciousness' (vol. 28, pp. 703–706), responding to the dispute over integrated information theory.
China's seven-department plan targets BCI breakthroughs by 2027 and 2–3 global leading firms by 2030
MIIT and six other departments/bodies issued Implementation Opinions on Promoting Innovative Development of the Brain-Computer Interface Industry (工信部联科〔2025〕164号). By 2027: breakthroughs in key BCI technologies, electrodes, chips and complete devices at international advanced level, and 2–3 industry clusters. By 2030: a secure and reliable industrial system and 2–3 globally influential leading firms. It frames BCI as coordinated development of life science and information science.
UK's ARIA backs a £69m Precision Neurotechnologies programme
ARIA's Precision Neurotechnologies programme, backed by £69m and within its Scalable Neural Interfaces opportunity space, aims to unlock new methods to interface with the human brain at the circuit level.
Japan's Moonshot Goal 1 runs an 'Internet of Brains' project toward a 2050 society 'free from limitations of body, brain, space, and time'
Japan's Moonshot R&D Goal 1 aims for a society by 2050 in which people are free from limitations of body, brain, space and time, using 'Cybernetic Avatar' technology. Its Kanai project develops avatars controlled via intention estimated from brain activity and brands itself the 'Internet of Brains' (IoB); on Aug 28, 2025 it announced a symposium titled 'Internet of Brains: A Blueprint for Connecting the Brain to the World'. JST's Kanai project page lists Sony Computer Science Laboratories researchers as R&D-item performers: Furuya Shinichi under Item [1] (IoB Interfaces) and Rekimoto Junichi under Item [2] (IoB Middleware), alongside Oizumi Masafumi (The University of Tokyo) under Item [2].
FDA approves Medtronic's adaptive (closed-loop) deep brain stimulation for Parkinson's
The FDA approved PMA supplement P960009/S478 on Feb 20, 2025 for adaptive DBS on Medtronic's Percept platform, which adjusts stimulation using the patient's own brain signals. Medtronic described it as the 'largest commercial launch of brain-computer interface technology ever' and said more than 40,000 patients have Percept devices (company claims).
FDA clears Magnus Medical's SAINT individualized TMS for depression
The FDA cleared the SAINT Neuromodulation System (K220177; letter dated Sept 1, 2022) for major depressive disorder in adults who have not improved on prior antidepressants. Magnus says SAINT combines advanced imaging, personalized targeting and novel stimulation patterns into individualized neurostimulation (company description).
Neuralink raises a $650 million Series E
Neuralink announced it closed a $650 million Series E with investors including ARK Invest, Founders Fund, G42, QIA, Sequoia Capital and Thrive Capital, saying the money will expand patient access and fund devices that 'deepen the connection between biological and artificial intelligence' (company announcement).
Meta FAIR decodes typed sentences from non-invasive brain recordings
Meta's FAIR lab, with the Basque Center on Cognition, Brain and Language, reported decoding up to 80% of characters typed by 35 healthy volunteers from MEG recordings, calling it a possible avenue for non-invasive BCIs and an 'important path to developing AMI' (advanced machine intelligence). Meta also announced a $2.2 million donation to the Rothschild Foundation Hospital.
Synchron unveils a roadmap to 'Chiral', a planned brain foundation model, with NVIDIA
On Mar 19, 2025 Synchron 'unveiled the roadmap to Chiral', which it describes as a foundation model of human cognition to be trained on de-identified neural data with NVIDIA; the release describes planned milestones, not a released model.
Google Research reports LLM speech representations align with brain activity in real conversations
Google Research described a Nature Human Behaviour study finding alignment between the internal representations of a Transformer speech-to-text model and the neural processing sequence in the human brain during real-life conversations.
IARPA-funded MICrONS maps a cubic millimetre of mouse cortex with function and wiring
The MICrONS consortium published in Nature a functional connectomics dataset co-registering activity recordings with an electron-microscopy reconstruction of more than 200,000 cells and 0.5 billion synapses in mouse visual cortex. The dataset page calls it the 'IARPA MICrONS dataset'.
FlyWire publishes the first complete wiring diagram of an adult fly brain
Dorkenwald et al. and the FlyWire Consortium published in Nature a whole-brain wiring diagram of an adult female fruit fly: 139,255 neurons and about 5 × 10^7 chemical synapses.
'State of Brain Emulation Report 2025' reassesses whole-brain emulation since the 2008 roadmap
Zanichelli, Schons, Freeman, Shiu and Arkhipov published a report reassessing brain-emulation progress since Sandberg and Bostrom's 2008 roadmap, organised around neural dynamics, connectomics and computational neuroscience.
Survey of 67 experts: median 90% that 'digital minds' are possible; 20% by 2030
Caviola and Saad surveyed 67 experts in early 2025 on 'digital minds' (computer systems capable of subjective experience). Medians: 90% possible in principle, 65% created by 2100, 20% by 2030.
Anthropic reports 'emergent introspective awareness' in Claude, without claiming consciousness
Anthropic published research on emergent introspective awareness in large language models and said its results do not tell whether Claude or any other AI system might be conscious.
Google and DeepMind researchers test whether LLMs trade points against stipulated pain, citing sentience debates
Researchers from Google's Paradigms of Intelligence team, Google DeepMind and the LSE (including Blaise Agüera y Arcas and Jonathan Birch) probed whether LLMs make trade-offs involving stipulated pain and pleasure, a question they say may bear on debates about LLM sentience, 'the capacity for valenced experiential states'.
NYU's Center for Mind, Ethics, and Policy becomes a permanent centre with a $6 million endowment
NYU's Center for Mind, Ethics, and Policy says 2024 was the year it became a permanent center with a $6 million endowment and released work on animal and AI welfare.
Largest study finds one in four unresponsive brain-injured patients show hidden cognitive responses
A New England Journal of Medicine study of 353 adults with disorders of consciousness found that, among 241 who could not respond visibly to commands, about one in four showed sustained cognitive responses on fMRI or EEG ('cognitive motor dissociation').
Australia lets authorised psychiatrists prescribe psilocybin and MDMA from July 2023
Australia's TGA announced that from 1 July 2023 authorised psychiatrists may prescribe MDMA for PTSD and psilocybin for treatment-resistant depression (Schedule 8 for those uses, Schedule 9 otherwise), with no TGA-evaluated approved product.
VA funds its first psychedelic-assisted therapy study since the 1960s
The VA announced about $1.5 million over five years for a study of MDMA-assisted therapy for veterans with PTSD and alcohol use disorder at Providence and West Haven VA medical centers, its first psychedelic-assisted therapy funding since the 1960s.
Delix reports positive phase Ib data for a non-hallucinogenic 'neuroplastogen' and FDA clearance for an at-home phase II
Delix Therapeutics said zalsupindole (DLX-001), which it calls a first-in-class non-hallucinogenic neuroplastogen, showed positive phase Ib efficacy data and that the FDA cleared a phase II design with at-home administration (company-reported).
Chile writes protection of 'brain activity' into its Constitution
Ley 21.383, published in Chile's Diario Oficial on Oct 25, 2021, amends Article 19 No. 1 of the Constitution so that scientific and technological development respects physical and psychic integrity, with the law to 'especially protect brain activity, as well as the information coming from it'.
Chile's Supreme Court orders Emotiv to delete a user's brain data
Chile's Supreme Court (rol 105.065-2023) held that the commercialisation and storage of brain-activity data from Emotiv's device affected the plaintiff's right to physical and psychic integrity and ordered Emotiv Inc. to delete the stored information.
Spain's Charter of Digital Rights includes rights over neurotechnologies
Spain's government adopted the Carta de Derechos Digitales, including digital rights in the use of neurotechnologies allowing law to regulate neurotechnology use to guarantee each person's control over their own identity, individual self-determination in decisions, confidentiality and control of data about their brain processes, and to regulate human-machine interfaces that could affect physical or psychic integrity.
OECD adopts the first international standard on responsible neurotechnology, defining 'cognitive liberty'
The OECD Council adopted the Recommendation on Responsible Innovation in Neurotechnology (OECD/LEGAL/0457), defining 'cognitive liberty' as 'the right to mental self-determination' and 'personal brain data', and setting principles that include safeguarding personal brain data (Principle 7) and avoiding hype (Principle 8c). Brazil adhered in 2022.
UN Human Rights Council adopts a resolution on neurotechnology and human rights
The Human Rights Council adopted resolution 51/3, 'Neurotechnology and human rights', on 6 October 2022.
Council of Europe's data-protection committee revises draft guidelines on neural data
The Convention 108 Consultative Committee issued revision 7 of its Draft Guidelines on Data Protection in the context of neuroscience and neurotechnologies (T-PD(2025)1rev7corr, 21 Oct 2026). In October 2025 a CDBIO-commissioned report analysed the European human-rights framework for neurotechnologies.
UN Human Rights Council asks for guidelines on neurotechnology and human rights
Resolution 58/6 'Neurotechnology and human rights' (A/HRC/58/L.9) was adopted without a vote on 2 April 2025; it asks the Advisory Committee to draft recommended guidelines for presentation at the Council's 64th session.
OECD publishes a neurodata-governance policy paper
OECD STI Policy Paper No. 195 maps neurodata flows in company practice and calls for clearer classification, consent and inferred-data guidance, and regulatory pathways for AI-enabled neurotechnology.
Texas enacts SB 2308 to fund FDA ibogaine trials
Texas Governor Greg Abbott signed SB 2308, effective immediately, letting the Health and Human Services Commission provide state matching funds for an FDA ibogaine drug trial run by a public university with a drug company and a hospital.
Texas says no drug-company proposal qualified for its ibogaine programme and will run its own research
A joint statement by the Lieutenant Governor and Speaker said $50 million had been allocated, HHSC found no drug-company proposal met requirements, and Texas would proceed with its own research programme.
FDA seeks public input on ibogaine trial design
FDA opened docket FDA-2026-N-10429 seeking input on early-phase ibogaine trial design, with comments due Nov 20, 2026; it said ARPA-H is funding early-phase trials and NIDA opioid-use-disorder research.
FDA grants Compass rolling NDA review and a Commissioner's National Priority Voucher
Compass Pathways said FDA granted its request for rolling review of the COMP360 psilocybin NDA and awarded a Commissioner's National Priority Voucher.
Anthropic keeps retired Claude Opus 3 available and acts on its 'retirement interview' preferences
Anthropic said it retired Claude Opus 3 on Jan 5, 2026 under its deprecation commitments, kept it available to paid users and by API request, and gave it a weekly essay channel after Opus 3 asked to share its 'musings and reflections'; it said it remains uncertain about the moral status of its models.
Anthropic reports functional emotion concepts inside Claude, without claiming feelings
Anthropic interpretability research identifies internal emotion-concept representations that influence model behaviour, stating that none of this tells us whether language models actually feel anything or have subjective experiences.
Anthropic reports a global-workspace-like mechanism in language models; it does not establish subjective consciousness
Anthropic published 'A global workspace in language models', interpretability research inspired by Global Workspace Theory, with external commentaries. In one commentary Neel Nanda (Google DeepMind) reports that he and MATS scholars Camila Blank and Agam Bhatia replicated the core claims on Qwen 3.6 27B; some evaluations (poetry, arithmetic) did not replicate.
Microsoft AI's CEO attacks Anthropic's model-welfare framing in 'Does Claude Have Rights?'
Mustafa Suleyman argued AI agents are not conscious, criticised Claude's constitution and the Opus 3 retirement interview as a training feedback loop, and promoted Microsoft's draft Humanist Superintelligence code of conduct.
Google DeepMind researcher argues AI can simulate but not instantiate consciousness
Alexander Lerchner's paper, listed on Google DeepMind's publications site, argues computational functionalism commits an 'abstraction fallacy' and that symbol manipulation cannot instantiate experience.
Merge Labs licenses Butterfly's ultrasound-on-chip for non-invasive BCIs
Butterfly Network announced a multi-year licensing agreement giving Merge Labs access to its Poseidon ultrasound-on-chip for ultrasound-based brain-computer interfaces, restating that Merge launched in Jan 2026 with $252M seed funding including OpenAI.
Synchron raises a $200M Series D
Synchron announced a $200M Series D led by Double Point Ventures, bringing total funding to $345M, and a cognitive-AI division.
Precision Neuroscience closes a $250M Series D
Precision said it closed a $250M Series D co-led by Pershing Square and the Ackman Oxman Institute, bringing total raised to $430M after more than 100 patient procedures at 18 institutions.
Paradromics completes its first long-term Connexus implant at University of Michigan Health
Paradromics said it completed its first long-term human implantation of the Connexus BCI at University of Michigan Health in June 2026.
Meta's Brain2Qwerty v2 reaches 61% word accuracy decoding typing from MEG
Meta FAIR reported Brain2Qwerty v2, trained on about 22,000 sentences from nine volunteers using MEG, at 61% average word accuracy (78% for the best participant), and released training code.
China issues a medical-device standard for AI-powered BCIs
China issued what state media called the world's first medical-device standard for AI-powered brain-computer interfaces, its third BCI device standard, effective Sept 1, 2027.
ARIA opens a £50m Massively Scalable Neurotechnologies programme
ARIA's programme call 'Massively Scalable Neurotechnologies for Human Health' (v1.2, 13 April 2026) says it is backed by £50 million.
Keio's Internet of Brains Society project enters a second phase on neurolaw
Keio's KGRI said its Moonshot Goal 1 IoB-S project moves into 'phase 2' from FY2026, aiming over five years to give guidance for domestic law and social rules on brain-machine interfaces and to build a basis for neurolaw.
Eon Systems says it combined a whole-brain fly model with a body simulation and aims to upload humans
Eon Systems wrote that it combined a Shiu-style whole-brain fly model with the NeuroMechFly body simulation and stated a goal to 'upload the first humans'.
Janelia, Cambridge and Google release a complete male fruit-fly CNS connectome
Janelia announced the complete connectome of a male fruit-fly central nervous system (brain and nerve cord, more than 166,000 neurons), with Google Research segmenting the images.
Nature publishes a densely reconstructed adult fly brain-and-cord connectome
Bates et al. report a connectome uniting the adult fly brain and ventral nerve cord and use it to study distributed control circuits.
Germany's Cyberagentur awards €30m to Zander Labs for neuroadaptive AI
Cyberagentur signed a contract worth 30 million euros with Zander Laboratories for its NAFAS neuroadaptivity programme (48 months).
Nature Neuroscience: adversarially trained AI derives consciousness markers validated across patients and animals
Toker et al. used adversarial AI on more than 680,000 samples and validated resulting markers on 565 patients, volunteers and animals.
Nature: largest single-site psilocybin brain-imaging study (62 adults)
Stoliker et al. report fMRI and EEG in 62 adults given psilocybin.
Focused ultrasound to the central thalamus reduces arousal in healthy volunteers
A Communications Biology study (Monti lab) reports central-thalamus sonication reduced arousal in 27 healthy volunteers.
Anthropic's usage policy adds a ban on sustained, needless cruelty toward its models (effective Nov 12, 2026)
On Oct 8, 2026 Anthropic announced a Usage Policy update, effective Nov 12, 2026, that adds 'a prohibition on sustained and needless abusive or cruel behavior toward our models'. Anthropic says it applies only in extreme cases and that Claude's ability to end such conversations 'will remain the primary enforcement mechanism'.
Preprint reports a linear 'pain direction' in 25 open-weight language models
A preprint by Tagliabue, Dung and Berg extracts a linear 'pain direction' from 25 open-weight models (2B-72B, 5 families) that separates pain from matched controls. First submitted to arXiv Sep 14, 2026; the current v3 was submitted Nov 3, 2026 and its manuscript is dated Oct 31, 2026.
Cortical Labs markets the CL1, which it calls 'the world's first code deployable biological computer'
Cortical Labs' product page offers the CL1 for purchase and describes real neurons cultivated on a silicon chip, running inside its 'Biological Intelligence Operating System (biOS)', to which users deploy code. Company-attributed record; launch date not established.
Counterpoint: NIH BRAIN Initiative funding fell from $680M (FY2023) to $321M (FY2025), then rose to $429M (FY2026)
NIH reports BRAIN Initiative totals of $680 million in FY2023, $402 million in FY2024, $321 million in FY2025 and $429 million in FY2026.