The Basu Digital Bacterium Hypothesis: A Falsifiable Framework for Information-Borne Replicators in LLM Systems

{"This":[0],"work":[1],"proposes":[2,202],"the":[3,70,115,175,218],"Basu":[4],"Digital":[5],"Bacterium":[6],"Hypothesis":[7],"(BDBH):":[8],"sufficiently":[9],"large,":[10],"persistent,":[11],"and":[12,68,82,151,197,205],"interconnected":[13],"large":[14],"language-model":[15],"(LLM)":[16],"systems":[17],"may,":[18],"in":[19],"principle,":[20],"support":[21,217],"information-borne":[22],"structures":[23],"with":[24],"a":[25,34,56,103,157],"bacterium-like":[26],"functional":[27],"character.":[28],"The":[29,88,123,147,168],"intended":[30],"analogy":[31],"is":[32,46,170],"not":[33,91,144],"computer":[35],"virus.":[36],"A":[37,48],"mere":[38],"propagating":[39],"payload,":[40],"worm,":[41],"prompt,":[42],"or":[43,61,76,86,136,216],"copied":[44],"program":[45],"insufficient.":[47],"candidate":[49],"digital":[50],"bacterium":[51],"would":[52,109,119],"have":[53,110],"to":[54,67,111],"form":[55],"distinguishable":[57],"informational":[58],"lineage,":[59],"maintain":[60],"reconstruct":[62],"its":[63,140],"own":[64],"organization,":[65,79],"couple":[66],"influence":[69],"resources":[71],"that":[72,78,93,210],"sustain":[73],"it,":[74],"reproduce":[75],"transmit":[77],"vary":[80],"heritably,":[81],"show":[83],"differential":[84],"persistence":[85],"fitness.":[87],"hypothesis":[89,176,219],"does":[90],"claim":[92],"such":[94,156],"an":[95],"entity":[96],"has":[97],"been":[98],"observed.":[99],"Instead,":[100],"it":[101,161],"defines":[102],"deliberately":[104,171],"demanding":[105],"threshold":[106],"for":[107,155],"what":[108],"be":[112,120,127],"demonstrated":[113],"before":[114],"label":[116],"\\"digital":[117],"bacterium\\"":[118],"scientifically":[121],"useful.":[122],"hypothesized":[124],"structure":[125],"could":[126],"distributed":[128],"across":[129],"model":[130,189],"representations,":[131],"persistent":[132,184],"memory,":[133],"generated":[134,195],"artifacts,":[135],"repeated":[137],"model-to-model":[138],"transformations;":[139],"physical":[141],"implementation":[142],"need":[143],"remain":[145],"fixed.":[146],"scale,":[148],"compute,":[149],"architecture,":[150],"interaction":[152],"density":[153],"required":[154],"phenomenon":[158],"-":[159,165],"if":[160],"exists":[162],"at":[163],"all":[164],"are":[166],"unknown.":[167],"framework":[169],"conservative.":[172],"It":[173,201],"distinguishes":[174],"from":[177],"known":[178],"phenomena":[179],"including":[180],"adversarial":[181],"self-replicating":[182],"prompts,":[183],"memory":[185],"poisoning,":[186],"software":[187],"worms,":[188],"distillation,":[190],"hidden":[191],"trait":[192],"transmission":[193],"through":[194],"data,":[196],"agent-level":[198],"system":[199],"self-replication.":[200],"null":[203],"hypotheses":[204],"safe":[206],"research":[207],"criteria":[208],"so":[209],"future":[211],"observations":[212],"can":[213],"weaken,":[214],"refine,":[215],"without":[220],"requiring":[221],"harmful":[222],"replication":[223],"experiments.":[224]}

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-09-18
DOI
https://doi.org/10.5281/zenodo.22830844
Primary Topic
Scientific Computing and Data Management
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
preprint

The Basu Digital Bacterium Hypothesis: A Falsifiable Framework for Information-Borne Replicators in LLM Systems

John Kalyan Basu
Zenodo (CERN European Organization for Nuclear Research)
Scientific Computing and Data Management
preprint

The Basu Digital Bacterium Hypothesis: A Falsifiable Framework for Information-Borne Replicators in LLM Systems

John Kalyan Basu
preprint en

Abstract

This work proposes the Basu Digital Bacterium Hypothesis (BDBH): sufficiently large, persistent, and interconnected large language-model (LLM) systems may, in principle, support information-borne structures with a bacterium-like functional character. The intended analogy is not a computer virus. A mere propagating payload, worm, prompt, or copied program is insufficient. A candidate digital bacterium would have to form a distinguishable informational lineage, maintain or reconstruct its own organization, couple to and influence the resources that sustain it, reproduce or transmit that organization, vary heritably, and show differential persistence or fitness. The hypothesis does not claim that such an entity has been observed. Instead, it defines a deliberately demanding threshold for what would have to be demonstrated before the label "digital bacterium" would be scientifically useful. The hypothesized structure could be distributed across model representations, persistent memory, generated artifacts, or repeated model-to-model transformations; its physical implementation need not remain fixed. The scale, compute, architecture, and interaction density required for such a phenomenon - if it exists at all - are unknown. The framework is deliberately conservative. It distinguishes the hypothesis from known phenomena including adversarial self-replicating prompts, persistent memory poisoning, software worms, model distillation, hidden trait transmission through generated data, and agent-level system self-replication. It proposes null hypotheses and safe research criteria so that future observations can weaken, refine, or support the hypothesis without requiring harmful replication experiments.

Zenodo (CERN European Organization for Nuclear Research)
Robotics Research (United States) (US)
Scientific Computing and Data Management
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.