iiinova — where your digital self keeps evolving

SecureBio, Inc. @securebio.org · 22/09/2026

Frontier AI now beats the best human virologists on biosecurity-relevant evals. Once an AI model bests every human expert we can test it against, a higher score stops telling us anything about real-world risk. Read our new blog post on how bio evals have to adapt: securebio.org/blog/evaluat...

Diagram showing that as AI model capabilities move from beginner to expert to superhuman, evaluations for chess and cybersecurity can shift from human-built proxies (puzzles, coding benchmarks) to real endpoints (real games, real exploits), while biosecurity cannot safely make that shift and must instead rely on indirect signals: experimental prediction, design tasks, defensive acceleration, and adjacent domains.

021

SecureBio, Inc. @securebio.org · 11/09/2026

Model accuracy dropped only 1.0–7.1 points on VCT-v2, and the gap between strongest and weakest models held steady, so past VCT results remain largely valid. VCT-v2 also has more headroom before saturation: at least ~30 points vs. ~23 on the original VCT.

Heatmap showing per-question performances of models and human experts on VCT-v2. Each cell represents the mean question × model accuracy, with darker colors indicating higher accuracy. Mean model accuracy on all questions is displayed to the right of the heatmap. Yellow cells represent data missing due to model refusals or missing expert baseline entries. The theoretical best model is taken to be the top-performing model on a per-question basis.

000

SecureBio, Inc. @securebio.org · 11/09/2026

We found that 27% of the original VCT questions were shortcut-exploitable (answerable without the image or question text). In all, we edited 163 questions (50.6%) and removed 43 questions (including 25 non-discriminating ones) to produce the 279-question VCT-v2 benchmark.

Flowchart showing the generation of the updated 279-question VCT-v2. Of the 322 VCT questions, 43 were removed, including 25 easy, non-discriminating questions; 163 were edited to improve their scientific accuracy, clarity, and/or resistance to heuristic shortcuts; and 116 remained unchanged.

100

SecureBio, Inc. @securebio.org · 11/09/2026

Top model accuracy on the original VCT has risen ~12 points since April 2025 (o3: 42.0%; GPT-5.6 Sol: 53.9%). To ensure this benchmark continues to capture the full dynamic range of model capabilities, we audited all 322 questions for validity and shortcut-resistance.

Benchmarks are restricted by the reliability limit of the answer key. This sets the “effective ceiling” of benchmarks and causes benchmark accuracy to stop scaling linearly with the scientific capability of the model beyond a certain threshold, resulting in lost dynamic range. Note: This schematic is an illustrative representation only.

100

SecureBio, Inc. @securebio.org · 12/08/2026

Meanwhile, our nasal swab sampling program, Zephyr, achieved a new single-day collection record (401 swabs) on May 9. Real-time data from this work is available on our Zephyr Dashboard. data.securebio.org/zephyr/

Screenshot of the Zephyr dashboard showing a bar graph with the number of samples collected per day, highlighting May 9th (401 samples).

100

SecureBio, Inc. @securebio.org · 12/08/2026

Lenni Justen (Sabeti lab) also published a preprint on normalization for wastewater metagenomic sequencing for quantitative pathogen tracking, identifying ways to make these data even more effective for monitoring trends in viral abundance and spread. www.medrxiv.org/content/10.6...

$From Justen et al. (2026): Wastewater metagenomic sequencing (WW-MGS) normalization marker abundance and dynamics. (b) A dot plot showing per-site median marker fractions; each point represents one site (n = 25), the horizontal black bar marks the across-site median, and the colors represent different geographic sites. (c) Pairwise within-site Pearson R on log10-transformed marker fractions. Each site's mean log10(marker) is subtracted before pooling, so correlations reflect sample-to-sample co-variation within sites. Establishing wastewater metagenomics as a quantitative pathogen monitoring tool with normalization Lennart Justen, Alessandro Zulli, Rose S. Kantor, Rebecca Y. Linfield, Leon S. Moskatel, Daniel Cunningham-Bryant, Jeff Kaufman, Marc C. Johnson, Michael R. McLaren, Pardis C. Sabeti doi: https://doi.org/10.64898/2026.07.14.26356442 https://www.medrxiv.org/content/10.64898/2026.07.14.26356442v1$

100

SecureBio, Inc. @securebio.org · 12/08/2026

Data from CASPER wastewater surveillance continue to yield key insights, like this wastewater RNA virome analysis from Rose Kantor (LLNL), Migun Shakya (LANL) et al. that may increase the efficiency and accuracy of future human pathogen screening efforts. www.medrxiv.org/content/10.6...

From Kantor et al. (2026): A bar graph showing the relative abundance of vOTUs from different viral orders in the wastewater virus genome database (WVDB). Bars are colored by the predicted host for each vOTU, and “other” includes vOTUs classified to families for which the ICTV virus properties table lists multiple hosts. Hosts are shown as “unknown” where no order- or family-level taxonomic call was made and the vOTU had no genus-level BLASTN hit against NCBI core-nt.

A genome-resolved view of the wastewater RNA virome
View ORCID ProfileRose S. Kantor, Migun Shakya, Nelson Ruth, Jason A. Rothman, Clayton Rushford, Devon A. Gregory, Aidan Epstein, Jeff T. Kaufman, Jonathan E. Allen, Patrick S. G. Chain, David H. O’Connor, Marc C. Johnson
doi: https://doi.org/10.64898/2026.05.19.26353600
https://www.medrxiv.org/content/10.64898/2026.05.19.26353600v2

100

SecureBio, Inc. @securebio.org · 09/07/2026

Sometimes our findings require action, but we see a lot of benign things too. Our systems that identify engineered pathogens can also find vaccine constructs or lab materials that pose no risk to the general population.

110

SecureBio, Inc. @securebio.org · 17/06/2026

The dashboard also includes a “Bio Capabilities Index,” an aggregate score underpinned by the same methodology as the Epoch Capabilities Index (ECI).

100

SecureBio, Inc. @securebio.org · 17/06/2026

Introducing SecureBio’s ‘Trends in Biology’ AI Benchmarks Dashboard SecureBio today releases a public dashboard with all of its AI model evaluation scores.

111

SecureBio, Inc. @securebio.org · 10/06/2026

As a final check to ensure our assessment truly tests real-world ability, we used code written by an LLM to instruct a liquid handling robot to carry out DNA assembly. And it worked.

200

SecureBio, Inc. @securebio.org · 13/04/2026

When SCV2 first started spreading it wouldn't have been in our databases yet, and the closest would have been bat SARS coronavirus ZC45, uploaded in early 2018. It aligns well to ZC45 along most of its genome, with the exception of the first half of the spike protein.

100

SecureBio, Inc. @securebio.org · 26/03/2026

We shared a preprint on our airplane lavatory study, collaborating with CDC and Ginkgo, where we found ~13x a higher human viral fraction. It's a preliminary result, but we're working to get access to additional airplane wastewater for further testing. www.medrxiv.org/content/10.6...

100

SecureBio, Inc. @securebio.org · 26/03/2026

With over 16,000 nasal swabs we've seen a wide variety of viruses, and using ONT sequencing we've recovered many near-complete genomes. These let us identify the specific strains spreading here.

100

SecureBio, Inc. @securebio.org · 26/03/2026

We also run project Zephyr, collecting nasal swabs out on the streets of Boston. We're averaging 600+ swabs weekly. You can read more in the Boston Globe: www.bostonglobe.com/2025/12/24/b...

100

SecureBio, Inc. @securebio.org · 26/03/2026

In November CASPER detected an unusual strain of mumps. Rich sequencing data helped us solve the puzzle: it was likely someone who got vaccinated shortly before flying to the US.

100

SecureBio, Inc. @securebio.org · 26/03/2026

We've continued to collaborate with our CASPER partners to sequence and analyze wastewater from across the US. We recently hit a significant milestone, where the majority of metagenomic wastewater sequencing data on SRA is from CASPER. www.medrxiv.org/content/10.6...

100

SecureBio, Inc. @securebio.org · 11/03/2026

This sequencing explains the seasonality of these viruses, the strains that are circulating, and the unique composition of wastewater samples by location. We demonstrate the viability of deep, untargeted wastewater metagenomics for tracking existing and novel human pathogens.

100

SecureBio, Inc. @securebio.org · 11/03/2026

We found that this sequencing closely tracks wastewater PCR and clinical data for SARS-CoV-2, influenza A, and RSV, while simultaneously capturing emerging infections like avian influenza, West Nile, and measles.

100

SecureBio, Inc. @securebio.org · 11/03/2026

“Deep untargeted wastewater metagenomic sequencing from sewersheds across the United States” represents half of all wastewater sequencing publicly available, and demonstrates the untapped potential of this technology.

100

SecureBio, Inc. @securebio.org · 11/03/2026

CASPER is a national wastewater surveillance program that collected and analyzed 1,206 samples covering more than 13 million people. For this preprint, a total of 1.2 trillion read pairs were generated, which represents the largest wastewater sequencing effort ever undertaken.

120

SecureBio, Inc. @securebio.org · 24/02/2026

Preprint: We applied untargeted viral metagenomics to composite airplane wastewater and found that human viruses had much higher relative abundance (median 13-fold) compared to municipal treatment-plant wastewater. This could substantially cut costs of using metagenomics to detect emerging viruses.

121

SecureBio, Inc. @securebio.org · 12/11/2025

We modeled how such a system could integrate metagenomic sequencing to detect both known and novel pathogens: naobservatory.org/blog/biothreat_radar

100

SecureBio, Inc. @securebio.org · 12/11/2025

Our paper on the sensitivity of wastewater metagenomic sequencing for early detection of viruses has now been published in The Lancet Microbe: www.thelancet.com/journals/lan...

100

SecureBio, Inc. @securebio.org · 12/11/2025

By now, our detection system frequently flags suspicious reads. But it doesn't recover the surrounding genome. Earlier this year, we showed how our outward assembly pipeline can recover these genomes, testing the algorithm on a flagged SARS-CoV-2/plasmid construct where it did extremely well.

100

SecureBio, Inc. @securebio.org · 12/11/2025

We’ve enhanced our data analysis systems, scaling metadata management, improving automated detection pipelines, and beginning integration of frontier LLMs for automated viral flag analysis.

100

SecureBio, Inc. @securebio.org · 12/11/2025

Zephyr, our Boston-based swab sampling program, has also expanded. With four dedicated field samplers working most weekdays, we collect 300-800 nasal swabs weekly. Viral reads are available via our dashboard: data.securebio.org/sampling-met...

100

SecureBio, Inc. @securebio.org · 12/11/2025

Nucleic Acid Observatory updates: - Major wastewater surveillance scaling: expanded to 31 sampling sites across 19 cities. - Zephyr swab program scaling, now 400-800 swabs per week. - New team members that lead response, wet-lab, and partnerships work.

110

SecureBio, Inc. @securebio.org · 25/07/2025

SecureBio’s NAO detection system flags suspicious reads, but doesn't recover the surrounding genome. We can recover these genomes with our outward assembly pipeline. We recently tested outward assembly on a flagged SARS-CoV-2/plasmid construct – it did extremely well!

100

SecureBio, Inc. @securebio.org · 10/07/2025

SARS-CoV-2 is highly detectable, but cold viruses are harder to detect. We will refine these estimates as we generate more sequencing data. naobservatory.org/blog/swab-ba...

100

SecureBio, Inc. @securebio.org · 10/07/2025

After our sampling sprint earlier this year, we’ve been scaling our Boston-based swab sampling program, hiring several field samplers. This has allowed us to collect 100-200 swabs per day. Over the coming quarter we will scale to additional weekdays and further optimize sampling.

110

SecureBio, Inc. @securebio.org · 10/07/2025

Our partners @solidevidence and Dave O'Connor have created new dashboards that present which pathogens we see in wastewater. Find the dashboard here: dholab.github.io/public_viz/

100

SecureBio, Inc. @securebio.org · 10/07/2025

We’ve generated another 487B read pairs, through both our own sequencing, and through Marc Johnson’s lab (@solidevidence). With the new data, our genetic engineering detection system has now analyzed 892B read pairs, flagging 413 chimeras, with 43 “benign positives”.

100

SecureBio, Inc. @securebio.org · 10/07/2025

Nucleic Acid Observatory updates: - We’ve further increased our sequencing capacity, producing 487B read pairs. - Our partners created dashboards that summarize which pathogens we routinely see in wastewater. - We’ve been scaling up our Boston-based swab sampling program.

230

SecureBio, Inc. @securebio.org · 26/06/2025

Linking this prevalence data with wastewater data, we estimated how easily different pathogens are detected in sequenced sewage. SARS-CoV-2 was again confirmed to be readily detectable, but cold viruses showed relatively low detectability.

100

SecureBio, Inc. @securebio.org · 26/06/2025

Using this we were able to estimate prevalence. We found that cold viruses are highly prevalent in winter: Many have a prevalence of 1% or higher!

100

SecureBio, Inc. @securebio.org · 26/06/2025

We created these estimates by analyzing nasal swabs, a recently added NAO sample stream. Collecting, pooling, and sequencing nasal swabs, we got a lot of information about the presence of cold viruses in the population, which we then linked with paired wastewater data.

110

SecureBio, Inc. @securebio.org · 26/06/2025

At the NAO, we’ve created new estimates on how well wastewater sequencing detects different pathogens. SARS-CoV-2 is again highly detectable, but common cold viruses are harder to detect. We will use this research to compare wastewater sequencing to other detection strategies.

3104

SecureBio, Inc. @securebio.org · 19/06/2025

Using this we were able to estimate prevalence. We found that cold viruses are highly prevalent in winter: Many have a prevalence of 1% or higher!

Chart of coronavirus and rhinovirus prevalence in the population

000

SecureBio, Inc. @securebio.org · 19/06/2025

Chart of 1) relative abundance of various viruses in wastewater and 2) positive swab pools of those same viruses

100

SecureBio, Inc. @securebio.org · 18/06/2025

SecureBio, partnering with Ginkgo Biosecurity will host an evening event on how to accelerate biosecurity on Thursday, July 31st, from 6:00-8:30 pm, in Boston (sign-up link in the next post).

110

SecureBio, Inc. @securebio.org · 25/04/2025

Standard metagenomic assembly at our scale would be prohibitively expensive. Instead, our approach searches billions of reads to find only those relevant to the seed, then assembles just those reads. 3/4

100

SecureBio, Inc. @securebio.org · 25/04/2025

We've now developed outward assembly—a new open-source pipeline that efficiently builds longer sequences by extending outward from a suspicious "seed" sequence, like a read section flagged by our chimera detection pipeline. 2/4

100

SecureBio, Inc. @securebio.org · 23/04/2025

Here is a sample of the results. Each dot is a set of questions. The y-axis is a model's score on that set minus the score of the virologist for which the set was tailored. Many models outperform most experts, with o3 exceeding 94% of them. 10/13

100

SecureBio, Inc. @securebio.org · 23/04/2025

Can AIs provide expert-level troubleshooting assistance for work with viruses? We built a new benchmark to answer that question. To our surprise, we found that leading models outperform the vast majority of practicing virologists we sampled. 🧵 1/13

110