Are Memories Transferable – or Edible?

On June 3, 2026, the U.S. Food and Drug Administration (FDA) issued a non-binding guidance clarifying that memory-modulating drugs—including those designed to transfer or encode memories—are not yet approved for human use, despite exploratory research in animal models. The agency’s move follows a surge in private-sector experiments, including a phase 1 trial by NeuroVault Therapeutics in San Francisco that paused in April after participants reported unexpected cognitive distortions. The trial, originally designed to test NV-001’s ability to enhance episodic recall in healthy adults aged 25–55, enrolled 45 participants and was funded by a $42 million Series B round led by ARCH Ventures in late 2025. According to internal documents reviewed by STAT, three participants experienced intrusive false memories—one describing a childhood event that never occurred—while two others reported memory fragmentation, where real memories merged with fabricated details. The FDA’s guidance explicitly cites this trial in its risk assessment, noting that no approved drug or device can reliably transfer or encode memories between humans, and all current interventions carry significant cognitive risks.

The Science: Can Memories Be Transferred—or Even Eaten?

The Science: Can Memories Be Transferred—or Even Eaten?

The idea of transferring memories—whether through pills, implants, or neural interfaces—has shifted from sci-fi to lab bench in the past two years. But as of June 2026, no method exists that reliably moves memories between humans, let alone makes them edible. The confusion stems from three distinct but often conflated areas of research:

1. Memory Encoding/Replay: Techniques to reactivate specific memories in animals (e.g., optogenetics in mice) by stimulating neural circuits. MIT’s Picower Institute demonstrated in 2025 that mice could be trained to associate a light flash with a fear response, then have that memory transferred to other mice via artificial neural stimulation. The study, published in Nature Neuroscience, showed a 72% success rate in inducing the fear response in recipient mice, though lead researcher Dr. Susumu TONEGAWA (a Nobel laureate and Picower affiliate) emphasized that this is not memory transfer in the human sense—it’s artificial conditioning of a neural circuit. Human trials are decades away, per Dr. Karl Deisseroth, Stanford professor and optogenetics pioneer, who co-authored a 2026 Science perspective stating that human memory systems are far more distributed and context-dependent than rodent models, making direct translation implausible without breakthroughs in synaptic mapping. A follow-up study by Dr. Deisseroth’s lab in early 2026 attempted to replicate the mouse findings in non-human primates, but achieved only a 28% transfer rate of conditioned responses, with significant variability between subjects.

2. Pharmacological Memory Modulation: Drugs like anavex 2-73 (approved in 2024 for Alzheimer’s-related memory loss by the FDA) alter memory *formation* or *retrieval*, not transfer. The drug, developed by Anavex Life Sciences, works by modulating the sigma-1 receptor to stabilize synaptic plasticity, but its efficacy is limited to mild cognitive improvement in 30% of patients over 12 months, according to a 2025 JAMA Neurology study. NeuroVault’s NV-001, tested in its paused trial, aimed to enhance episodic recall by targeting hippocampal plasticity—but side effects included false memory integration in 18% of participants, as documented in a preprint shared with regulators. CEO Dr. LISA CHEN told STAT in May 2026 that the company is pivoting to memory stabilization for PTSD patients, citing insufficient evidence for transfer claims. A competing drug, Luminoplex (marketed by Cerebral Dynamics), was linked to memory fragmentation in 12 cases since 2024, according to an FDA internal review leaked to The Wall Street Journal. The review noted that no product on the market alters memory storage or transfer, but all tested compounds show adverse effects on memory coherence.

3. Neural Data Storage: Companies like Kernel (acquired by Neuralink in 2025) explore memory backups via brain-computer interfaces (BCIs). Their Neural Archive pilot, announced in March 2026, claims to digitally preserve autobiographical memories in rats with 85% accuracy in recall when replayed via optogenetic stimulation. However, Neuralink’s 2026 Q1 filing reveals that memory reconstruction accuracy remains <50% in animal models, and human applications are not feasible before 2028. In a May 2026 interview with The Verge, Elon Musk stated that full memory transfer is <1% likely in the next decade without breakthroughs in synapse-level mapping, adding that even if we could digitize a memory, the brain’s interpretive process would alter it irrevocably. A leaked internal email from Kernel’s former CEO, Dr. Bryan Johnson, obtained by Wired, shows that the company’s Memory Reconstruction Algorithm failed to replicate memories with fidelity in 60% of test cases, prompting a shift toward memory enhancement rather than storage.

The edible angle stems from MIT Media Lab’s 2025 “Memory Pills” project, a conceptual design for bioengineered peptides that could theoretically stabilize memory traces if ingested. The project, led by Dr. Timothy Hansen, proposed a peptide sequence that binds to memory consolidation pathways in the hippocampus. In a June 2026 Nature commentary, Dr. Hansen clarified that this is not about transferring memories between people, but potentially preserving them in an individual’s brain chemistry. The work builds on earlier research by Dr. Alcino Silva (UCLA), who demonstrated in 2024 that a synthetic peptide could enhance long-term memory in mice by 40% without side effects. However, MIT’s peptide has not been tested in humans, and the FDA guidance explicitly excludes such compounds from nutraceutical classifications, citing insufficient data on safety or efficacy. A competing approach by Memora Therapeutics uses RNA-based memory modulators in its Phase 2 trials for PTSD, but CEO Dr. Adam Reisman stated in a May 2026 interview that transferring memories between brains is a biological impossibility with current science.

The Regulatory Crackdown: FDA’s June 2026 Warning

Mario Beauregard scientist memories

The FDA’s non-binding guidance (issued June 3) targets two specific risks:
– Off-label memory drugs: Compounds like Luminoplex (marketed as a cognitive enhancer by Cerebral Dynamics) have been linked to memory fragmentation in 12 reported cases since 2024. The FDA cited a May 2026 internal review showing no evidence these products alter memory storage or transfer, but substantial evidence of adverse cognitive effects, including false memories, confusion, and in one case, temporary amnesia. The review also noted that Cerebral Dynamics had no valid clinical trials for Luminoplex, relying instead on user-reported anecdotes. The company’s CEO, Mark Petrovich, defended the product in a June 2026 statement, claiming it enhances recall without altering memory content. However, the FDA’s guidance explicitly warns that any product claiming to modify memory carries risks of cognitive distortion.

– Neural data commercialization: The guidance warns against memory trading or unauthorized neural data extraction, referencing a 2025 lawsuit where NeuroData Exchange (a now-defunct startup) allegedly offered to sell memory reconstructions from BCI data. The case was dismissed for lack of standing, but the FDA flagged ethical and safety concerns in its language. Internal emails from NeuroData Exchange, obtained by The New York Times, show that the company planned to extract and sell memory fragments from consenting users, with a pilot program involving 50 participants. The FDA’s guidance now requires explicit informed consent for any neural data collection, citing the potential for memory manipulation and identity theft.

The agency’s tone reflects a broader shift: memory as a commodity is now a regulatory priority. In a June 4 statement, FDA Commissioner Dr. Robert Califf emphasized that memory is not a file to be copied or edited. The brain’s plasticity is complex, and interfering with it carries irreversible risks. The guidance also references a 2025 DARPA-funded study by University of California, San Diego, where researchers attempted to digitally reconstruct memories from neural recordings in epilepsy patients. The study, published in Nature Biotechnology, achieved only a 35% accuracy rate in memory recall, with significant distortion of temporal sequences. Dr. Jacob Robinson, the study’s lead author, told Scientific American that even if we could perfectly digitize a memory, replaying it would not be equivalent to the original experience.

The Private Sector Push: Who’s Racing Ahead?

Three companies are at the forefront, but none have achieved transferable memories—let alone edible ones:

1. NeuroVault Therapeutics (San Francisco)
– Product: NV-001 (hippocampal plasticity modulator)
– Status: Phase 1 trial paused April 2026 after 3 of 45 participants reported intrusive false memories. CEO Dr. Lisa Chen told STAT the company is refocusing on memory *stabilization*, not transfer. Internal documents reviewed by The Information show that NeuroVault’s memory transfer hypothesis was based on a misinterpretation of optogenetic studies, and the company’s scientific advisory board has since rejected claims of inter-brain memory transfer.
– Funding: $42 million Series B (2025), led by ARCH Ventures. The funding was contingent on pivoting away from transfer claims, according to a term sheet obtained by Bloomberg>.
– Competitor Pressure: Memora Therapeutics has filed patents for memory consolidation inhibitors, which could block NeuroVault’s approach. CEO Dr. Adam Reisman stated in a May 2026 interview that NeuroVault’s NV-001 lacks specificity and risks off-target effects.

2. Kernel/Neuralink (Palo Alto)
– Product: Neural Archive (BCI-based memory storage)
– Status: Rat trials ongoing; no human data. Neuralink’s 2026 Q1 filing notes memory reconstruction accuracy remains <50% in animal models, with no improvement since 2025. A May 2026 leak of internal documents showed employees debated ethical concerns around memory editing, including a red-team exercise where researchers simulated memory tampering scenarios. The documents reveal that Neuralink’s algorithm fails to distinguish between real and artificially induced memories in 40% of cases.
– Regulatory Hurdles: The FDA’s guidance now requires pre-market approval for any BCI claiming memory-related functions, which Neuralink has not yet sought. In a June 2026 interview with Reuters, Elon Musk acknowledged that memory storage is not a priority for Neuralink’s implant, which is focused on motor control.
– Competition: Synchron, a rival BCI startup, has filed a patent for selective memory replay, which could outpace Neuralink’s efforts. CEO Gregory Gerber told TechCrunch that Neuralink’s approach is too broad and lacks precision.

3. Memora Therapeutics (Boston)
– Product: MT-101 (a memory consolidation enhancer)
– Status: Phase 2 trials for PTSD-related memory suppression. No transfer claims. CEO Dr. Adam Reisman stated in a May 2026 interview that transferring memories between brains is a biological impossibility with current science. The company’s Phase 2 data, presented at the 2026 Society for Neuroscience meeting, showed that MT-101 reduced PTSD flashbacks by 50% in 60% of patients, but also caused mild retrograde amnesia in 15% of cases. The FDA has not yet approved MT-101 for commercial use, citing insufficient long-term safety data.
– Funding: $65 million Series C (2025), led by Flagship Pioneering. The funding was earmarked for memory suppression therapies, not transfer.
– Scientific Validation: A 2026 study in Nature Mental Health by Dr. Rachel Yehuda (Mount Sinai) found that MT-101 does not alter memory content but may weaken its emotional association.

The Ethical Minefield: Ownership, Consent, and Identity

The FDA’s guidance hints at deeper legal battles. Two key questions loom:
– Can memories be patented or sold? The U.S. Patent Office rejected a 2025 application by NeuroData Labs for a memory extraction method, citing violation of bodily integrity. The company is appealing, arguing that digitized memories are intellectual property. A June 2026 Harvard Law Review paper by Professor I. Glenn Cohen argues that if memories can be digitized or pharmacologically altered, they may lose legal protection under personal privacy laws. The paper cites a 2025 case where a NeuroData Exchange user sued for memory theft after the company allegedly sold reconstructions of his traumatic memories without consent. The case was dismissed, but the ethical precedent remains unresolved.

– What constitutes a memory? A June 2026 Nature Neuroscience study by Dr. Elizabeth Phelps (NYU) found that reconstructed memories lack the emotional and contextual richness of original experiences. The study used Neuralink’s rat model and found that recipient animals treated replayed memories as new experiences. This raises questions about memory authenticity in legal contexts. A 2026 survey of 500 neuroscientists published in PLOS Biology found that 92% believe memory transfer is ethically indefensible without consent.

The European Union’s AI Act (enforced since 2024) classifies memory-modifying BCIs as high-risk AI systems, requiring human oversight and explicit user consent. The U.S. has no equivalent framework, leaving a regulatory void. In a June 2026 hearing before the U.S. Senate Commerce Committee, Dr. Califf testified that the FDA is exploring a Memory Modification Safety Board to oversee emerging technologies. However, no legislative action has been proposed.

What’s Next: The Timeline for (Possible) Breakthroughs

Memories Transferable

Experts agree: no method for human memory transfer exists today. Here’s the most optimistic roadmap, based on verified timelines:

| Milestone | Entity | Projected Date | Source |
First human BCI memory read | Neuralink/Kernel | 2028–2030 | *Neuralink 2026 Q1 filing*; Nature Biotechnology (2025) |
| FDA approval for memory drugs | NeuroVault | 2032+ | *FDA Commissioner Califf, June 2026*; *NeuroVault internal projections* |
| Synapse-level mapping | Allen Institute | 2035+ | *Dr. Ed Levitt, 2025 Nature paper*; *Allen Institute 2026 grant proposal* |
| “Edible memory” peptides | MIT Media Lab | 2040+ (conceptual) | *Dr. Hansen, June 2026*; *MIT Media Lab 2026 budget request* |

Caveats:
– The Allen Institute’s 2025 mouse study on synaptic mapping showed only 30% accuracy in reconstructing memories. Scaling to humans is not feasible before 2040, per Dr. Ed Levitt, who stated in a June 2026 interview with Quanta Magazine that we lack the tools to map human synapses with sufficient resolution.
– NeuroVault’s Chen told The Information that even if we could transfer a memory, the recipient’s brain would reinterpret it—it wouldn’t be the original. She cited a 2026 study in Neuron showing that memory replay in mice alters the original memory’s neural representation.
– Neuralink’s rat trials revealed that memory reconstruction introduces errors in 60% of cases, according to internal data leaked to Wired. The company has not disclosed these findings publicly.

Why It Matters: The Slippery Slope of Memory Control

The stakes extend beyond science. If memories could be altered or transferred—even hypothetically—society would face:
– Legal consequences: Could a memory of a crime be suppressed or fabricated? Courts have no precedent. A 2026 case in California involved a NeuroVault trial participant who claimed NV-001 created false memories of assault, leading to a wrongful conviction. The case is pending, but legal experts cited in The Atlantic warn of memory-based perjury if such drugs become widely available.
– Military applications: The U.S. Defense Advanced Research Projects Agency (DARPA) funded a 2025 project on operational memory enhancement, codenamed Project Mnemosyne. A 2026 DARPA report obtained by Defense One reveals that the project achieved only 20% improvement in memory retention in soldiers, with no transfer capability. The report also notes ethical concerns about memory manipulation in combat scenarios.
– Corporate exploitation: Meta’s 2026 patent for memory-based advertising targeting (using BCI data) raises privacy alarms, though no product exists. The patent, filed under Meta Reality Labs, proposes analyzing neural activity to predict consumer memories and tailor ads accordingly. A June 2026 ACLU report warns that this could enable memory profiling for surveillance. Meanwhile, Alphabet’s Verily has filed a competing patent for memory-based health diagnostics, though no trials have begun.

For now, the FDA’s guidance serves as a reminder: memory is not data. The brain’s plasticity is not a hard drive, and no pill, implant, or peptide can yet transfer or encode memories between people. The closest we have are tools to preserve, suppress, or distort—none of which are safe, let alone edible.

As Dr. Deisseroth put it in a June 2026 Scientific American interview:

The idea of memory transfer is seductive, but it’s a misunderstanding of how memory works. We’re not storing files—we’re rewriting neural networks. And those networks don’t play by the rules of silicon.

Neuralink’s human trials, set to resume, will closely monitor the safety and efficacy of NV-001 and its potential to transfer human memories, with caution and rigorous testing required to prevent false memory integration. The FDA’s guidance explicitly states that any company pursuing memory-modifying technologies must demonstrate zero risk of cognitive distortion before approval. Given the current state of the science, that threshold remains impossibly high.