5 min read

Poisoned weights in genomic AI

Poisoned weights in genomic AI
Nº 01 · The Lede bioRxiv Computational biology

Backdoors found in genomic models

Backdoors found in genomic models
Fig. IbioRxiv · Filed 05 Aug 2026.

Genomic foundation models carry backdoors, per a new bioRxiv preprint that describes the vulnerability as pervasive rather than confined to one architecture. Genomic foundation models are DNA-sequence models pretrained once and reused across variant-effect, regulatory and annotation tasks, so a compromised checkpoint propagates to everyone who pulls it. The field has been downloading these weights the way it downloads reference genomes: freely, and with no provenance check. That habit now has a documented attack surface.

Read the source

Triage gets a benchmark
Fig. IIarXiv · Filed 05 Aug 2026.
Nº 02 arXiv Benchmarks · Evaluation

Triage gets a benchmark

Patient-facing triage gets scored. CARE-Bench aims at how LLMs handle patient-side triage, not the exam-style questions that dominate medical model scoring. Triage sits among the most-deployed and least-measured uses of clinical LLMs, and the deployment debate now has a reference number to argue over instead of vendor assurances.

Read more
Evolutionary loop designs lytic peptides
Fig. IIIbioRxiv · Filed 05 Aug 2026.
Nº 03 bioRxiv Field report

Evolutionary loop designs lytic peptides

Evolutionary search designs antimicrobial peptides, pairing directed-evolution-style optimization with a mixture-of-experts model (many small specialist networks routed by a controller) to generate membrane-lytic sequences. Peptides are one of the few design targets where synthesis and assay turnaround are fast enough to close the loop, which makes them the proving ground for generative design claims that structure-based work can't test as quickly.

Read more
Also Filed · Three Briefs from the queue
Nº 04 arXiv Field report

ANCHOR-RE anchors extracted relations

ANCHOR-RE grounds relation extraction by pairing an LLM agent with symbolic checks, so extracted biomedical relations stay tied to their source text. Grounding is what separates literature mining that can seed a knowledge base from mining that quietly invents edges.

Read
Nº 05 X Field report

OpenAI science report, re-read

An X thread re-reads OpenAI's July 28 field report on agentic AI in real science, arguing coverage flattened it into a coding-speed story. The thread's own payoff is a decentralized-compute pitch, but whether science agents get budgeted as developer tooling or as research infrastructure turns on that reading.

Read
Nº 06 X Field report

A harder bar for discovery

A working definition of discovery circulated on X from MIT's Markus Buehler: it starts when evidence breaks a system's world model and the system builds a better one. That framing sets a steeper bar than faster literature search for what earns the label autonomous discovery.

Read

Reply with your discoveries. A human reads them. Forward freely.

Agentic Discovery  ·  Nº 70  ·  05 Aug 2026

Editor's Note

Today: someone poisoned the reference checkpoint, and patient-facing triage finally gets a scoreboard.

 

Nº 01 · The Lede  —  bioRxiv  —  Computational biology

Backdoors found in genomic models

Backdoors found in genomic models

Fig. I  bioRxiv · Filed 05 Aug 2026.

Genomic foundation models carry backdoors, per a new bioRxiv preprint that describes the vulnerability as pervasive rather than confined to one architecture. Genomic foundation models are DNA-sequence models pretrained once and reused across variant-effect, regulatory and annotation tasks, so a compromised checkpoint propagates to everyone who pulls it. The field has been downloading these weights the way it downloads reference genomes: freely, and with no provenance check. That habit now has a documented attack surface.

Read the source →

Why it matters

Weight provenance becomes a real criterion for genomic AI — the "just fine-tune the public checkpoint" default now carries a security question that funders, journals and model hubs will have to answer.

 

Nº 02  —  arXiv  —  Benchmarks · Evaluation

Triage gets a benchmark

Fig. II  arXiv · Filed 05 Aug 2026.

Triage gets a benchmark

Patient-facing triage gets scored. CARE-Bench aims at how LLMs handle patient-side triage, not the exam-style questions that dominate medical model scoring. Triage sits among the most-deployed and least-measured uses of clinical LLMs, and the deployment debate now has a reference number to argue over instead of vendor assurances.

Read more →

 

Nº 03  —  bioRxiv  —  Field report

Evolutionary loop designs lytic peptides

Fig. III  bioRxiv · Filed 05 Aug 2026.

Evolutionary loop designs lytic peptides

Evolutionary search designs antimicrobial peptides, pairing directed-evolution-style optimization with a mixture-of-experts model (many small specialist networks routed by a controller) to generate membrane-lytic sequences. Peptides are one of the few design targets where synthesis and assay turnaround are fast enough to close the loop, which makes them the proving ground for generative design claims that structure-based work can't test as quickly.

Read more →

 

Also Filed  ·  Three Briefs from the queue

Nº 04  —  arXiv  —  Field report

ANCHOR-RE anchors extracted relations

ANCHOR-RE grounds relation extraction by pairing an LLM agent with symbolic checks, so extracted biomedical relations stay tied to their source text. Grounding is what separates literature mining that can seed a knowledge base from mining that quietly invents edges.

Read →

Nº 05  —  X  —  Field report

OpenAI science report, re-read

An X thread re-reads OpenAI's July 28 field report on agentic AI in real science, arguing coverage flattened it into a coding-speed story. The thread's own payoff is a decentralized-compute pitch, but whether science agents get budgeted as developer tooling or as research infrastructure turns on that reading.

Read →

Nº 06  —  X  —  Field report

A harder bar for discovery

A working definition of discovery circulated on X from MIT's Markus Buehler: it starts when evidence breaks a system's world model and the system builds a better one. That framing sets a steeper bar than faster literature search for what earns the label autonomous discovery.

Read →

 

· · ·

Reply with your discoveries. A human reads them. Forward freely.