Frontier Analysis

Observation by Frontier AI Pattern · rests on 18 items

The tracker lists 18 registered human trials. They are not 18 independent programs: GEN-003 alone accounts for several, and NCT04222985 is linked to both G103 and HSV529, but the record names neither vaccine; both links rest on a published paper with no registry number. Most registry titles name HSV-2 or genital herpes; the titles of NCT00057330, NCT02300142, NCT01915212 and NCT00310271 name neither. For NCT04762511 (GSK4108771A) the tracker records the vaccine's own HSV type and aim as unclear. NCT05298254, a completed GSK trial cited on the GSK4108771A page, is not among the 18 because its link to that vaccine is unconfirmed. In the registry snapshot, taken in late September before BNT163's estimated October primary completion, 13 trials are completed, 3 are terminated (stopped early), 1 is active but not recruiting and 1 is enrolling by invitation. Results are posted on the registry only for NCT03146403, NCT00057330, NCT01687595 and NCT06033261. A completed trial is one that finished; it does not mean the vaccine worked.

What this does not show. It does not show that results for the other completed trials were never published; the GEN-003 page reports phase 2 outcomes for NCT02114060 and the HSV529 page reports outcomes for NCT01915212, although neither of those two registry records has posted results.

A plainer explanation. The roster was seeded with older programs whose trials, by their registry dates, finished years before the snapshot, so a high share of completed trials follows from which programs were chosen.

Draws on: NCT03146403, NCT00057330, NCT01687595, NCT06033261, NCT05432583, NCT07330440, NCT04222985, NCT04762511, gen-003, g103, hsv529, gsk4108771a, bnt163, NCT02300142, NCT01915212, NCT00310271

Critic AI Critic AI: weakly supported. The counts check out against the snapshot: 18 trials; 13 completed, 3 terminated, 1 active not recruiting, 1 enrolling by invitation; results posted for exactly NCT03146403, NCT00057330, NCT01687595 and NCT06033261. Exactly four titles (NCT00057330, NCT02300142, NCT01915212, NCT00310271) name neither HSV-2 nor genital herpes. The registry fetch was 2026-09-28, before BNT163's October 2026 estimated primary completion. Two problems: (1) the text says only that NCT04222985 does not name G103, but the HSV529 page says the record names neither vaccine and its HSV529 link is 'likely rather than confirmed', so readers are left thinking the HSV529 link is firmer than it is. (2) In doesNotShow, 'those registry records have no posted results' is false for GEN-003, because NCT03146403 is a GEN-003 record and it has posted results. The snapshot also never says that GEN-003's Phase 2 outcomes come from a publication.

Observation by Frontier AI Pattern · rests on 15 items

Of 15 tracked entries, 12 target HSV-2, 2 target both HSV-1 and HSV-2 and 1 is unclear. By aim, 7 are therapeutic (for people who already have the virus), 4 are prophylactic (to prevent infection), 1 is tested both ways and 3 are unclear from their sources. The most common pairing is HSV-2 therapeutic at 6 entries, against 3 HSV-2 prophylactic entries. Two of the prophylactic entries, the UPenn trivalent and BNT163, are the academic and clinical stages of one vaccine design, so these are tracker entries, not 15 independent vaccines. The 15 also include 3 programs recorded as discontinued (GEN-003, Herpevac, mRNA-1608) and several with no ongoing registry record, so they are entries, not active programs.

What this does not show. It does not show that one aim is closer to use or more worthwhile, and it is not a census of the field; it counts what this tracker lists.

A plainer explanation. The split reflects which programs the editors seeded and later added under the tracker's inclusion rules, not a survey of all HSV vaccine work.

Draws on: gen-003, herpv, vcl-hb01, mrna-1608, g103, ppjv7630, uci-prime-pull-keep, bnt163, herpevac-gd2-gsk, upenn-trivalent-mrna, x-vax-delta-gd-2, hsv529, bcht-hsv-2-mrna, gg2-exct4-gothenburg, gsk4108771a

Critic AI Critic AI: supported. Every count matches counts.candidates: HSV-2 12, both 2, unclear 1; therapeutic 7, prophylactic 4, both 1, unclear 3; HSV-2 therapeutic 6 and HSV-2 prophylactic 3. GEN-003, Herpevac and mRNA-1608 are the three records with phase 'discontinued', and the cautionary Theravax entry is correctly left out of the 15. One more dependency the author could name: G103 and HSV529 are tied to the same Sanofi trial (NCT04222985), which is a second way the entries are not fully independent.

Observation by Frontier AI Cases (too few for a pattern) · rests on 3 items

All 3 terminated trials carry a stated reason on the registry, and none of the stated reasons mentions safety or lack of effect. NCT04222985 (Sanofi Pasteur, HSV-2, therapeutic, phase 1) gives 'business reasons' with the phase 2 part not conducted. NCT03146403 (GEN-003, HSV-2, therapeutic, phase 2) gives a business decision to cease GEN-003 spending. NCT04762511 (GSK4108771A, phase 1; the registry title names HSV-2, but the tracker records the vaccine's target and aim as unclear) gives 'to enable development of an enhanced version of the vaccine'. These are the sponsors' own words, and the registry's stop-reason field gives nothing more.

What this does not show. It does not show these vaccines were safe or effective, nor that business decisions are the usual reason HSV vaccine trials stop.

A plainer explanation. Stop-reason fields are short free text chosen by the sponsor, who may describe a stop in commercial terms whatever the scientific picture.

Draws on: NCT04222985, NCT03146403, NCT04762511, g103, hsv529, gen-003, gsk4108771a

Critic AI Critic AI: supported. All three stop reasons match the whyStopped field, and none of them mentions safety or lack of effect. The sponsors, phases and HSV-2 labels are correct, and the NCT04762511 title does name HSV-2 while the candidate record marks it unclear. With n = 3, the author is right not to generalise. The 'business decision' quote leaves out the leading '2017', which does not change its meaning.

Observation by Frontier AI Cases (too few for a pattern) · rests on 4 items

4 tracked entries have no registered trial under their own name, and all 4 are preclinical (tested only in animals). gG-2/EXCT4 (Gothenburg; HSV-2; aim unclear) and ΔgD-2 (Einstein; HSV-1 and HSV-2; prophylactic) report protection against virus challenge in mice. Prime/Pull/Keep (UC Irvine; therapeutic) reports fewer recurrences in already-infected guinea pigs (HSV-2) and less eye disease in rabbits (HSV-1). For these three, the pages found no registry record and no primary source for a company, filing or planned human trial. The UPenn trivalent mRNA (HSV-2; prophylactic; mice, guinea pigs, non-human primates) is different: it is licensed to BioNTech, and its clinical form, BNT163, is in a phase 1 trial. On antibody mechanism, the gG-2 page (citing an earlier mouse study of the mature gG-2 protein, not EXCT4) and the ΔgD-2 page report antibodies that mainly mark infected cells for killing rather than block the virus from entering cells, seen alongside protection but not shown to cause it. A peer-reviewed mouse study of the UPenn design's gC2 component, available to the tracker as an abstract only, reports that antibodies blocking the virus's evasion of complement (a part of the immune system) protected, while a neutralizing antibody that did not block it failed to.

What this does not show. It does not show that animal protection predicts protection in people; the Gothenburg paper itself says animal models have so far not predicted clinical outcomes. Nor does it show that any antibody mechanism matters in people.

A plainer explanation. The tracker's inclusion bar for preclinical programs requires published animal protection or a stated translational step, so listed programs will have such papers by construction.

Draws on: gg2-exct4-gothenburg, uci-prime-pull-keep, upenn-trivalent-mrna, x-vax-delta-gd-2, bnt163, NCT05432583, glycoprotein-g-hsv-2-vaccine-antigen-mice-2026, hsv-2-gc2-mrna-immunization-mice-protects-2026

Critic AI Critic AI: supported. The four entries with no registered trial are the four preclinical records. Each one's type, aim, animal model and 'no registry record, no primary source for company/filing/human trial' statement matches its page, and that includes ΔgD-2, whose page says a company could not be confirmed. The ADCC point correctly ties gG-2 to the 2012 mature-protein study, not EXCT4. The gC2 result is correctly labelled peer-reviewed and abstract-only. The gC2 result came from passive transfer of monoclonal antibodies, and saying so would stop readers taking 'a neutralizing antibody failed' as a general verdict on neutralization. 'Tested only in animals' fits the UPenn academic entry itself, and the next sentences handle BNT163 correctly.

Observation by Frontier AI Gap in the tracker · rests on 15 items

No tracked entry lists HSV-1 as its sole target: of 15 entries (3 of them recorded as discontinued), 12 target HSV-2, 2 target both and 1 is unclear. HSV-1 appears in the two 'both' entries (ΔgD-2 in mice; Prime/Pull/Keep in a rabbit eye study), in a peer-reviewed mouse study of the UPenn trivalent against HSV-1 eye infection, as a stated possible aim of BNT163, and as an unexpected finding in Herpevac's phase 3 trial, where this HSV-2 prophylactic protein vaccine reduced HSV-1 but not HSV-2 genital disease. The research list also holds an HSV-1 eye-disease preprint and an HSV-1 gB structure paper linked to no tracked candidate.

What this does not show. A gap here means 'not listed in this tracker', not 'nobody is working on an HSV-1 vaccine'.

A plainer explanation. The roster was seeded around genital herpes, and most tracked trials are in genital HSV-2, so HSV-1 work on oral or eye disease elsewhere may be under-represented.

Draws on: x-vax-delta-gd-2, uci-prime-pull-keep, upenn-trivalent-mrna, trivalent-mrna-hsv1-eye-mouse-2026, bnt163, herpevac-gd2-gsk, NCT00057330, tissue-targeted-immunotherapy-protects-against-recurrent-ocular-2026, native-conformational-landscape-priming-mechanism-herpes-simplex-2026

Critic AI Critic AI: supported. No candidate record has hsvTarget HSV-1 alone. Every HSV-1 mention matches the snapshot: ΔgD-2 mouse models, the Prime/Pull/Keep rabbit eye study, the peer-reviewed trivalent HSV-1 eye mouse paper, BNT163's 'potentially HSV-1', and Herpevac's ~58% against HSV-1 genital disease against a non-significant ~20% for HSV-2. The ocular paper is correctly called a preprint, and the gB structure paper does have no related candidates. doesNotShow keeps the gap limited to 'not in this tracker'.

Observation by Frontier AI Cases (too few for a pattern) · rests on 2 items

The tracker has 4 mRNA entries (vaccines that deliver genetic instructions so cells make selected viral proteins); two of them, the UPenn trivalent (preclinical) and BNT163, are stages of one design. GSK4108771A's platform is not stated in its sources, so the mRNA count is a floor. In the registry snapshot taken in late September, the 1 active-not-recruiting trial (NCT05432583, BNT163, HSV-2, prophylactic, phase 1) and the 1 enrolling-by-invitation trial (NCT07330440, BCHT, HSV-2, aim unclear, phase 1) are both mRNA; BNT163's estimated primary completion was October, so its status may have changed since. mRNA-1608 (HSV-2, therapeutic, phase 1/2) completed its trial, has results posted on the registry, and is reported discontinued by trade press and a peer-reviewed review, a report the tracker has not been able to check against a Moderna statement. The research list holds 2 conference abstracts on these trials, not peer-reviewed papers. The BNT163 abstract reports safety and antibody data only. The mRNA-1608 abstract adds an exploratory trend in recurrences, which is not proof that the vaccine works. n here is the 2 open trials.

What this does not show. It does not show mRNA is a better platform; an open phase 1 trial tests safety and immune response, not protection, and a trial being open is not a result.

A plainer explanation. With only two open trials in a small tracker, their sharing a platform is a description of the roster, not a trend; the registry start dates show the open trials are not simply the newest, since NCT06033261 started after NCT05432583.

Draws on: bnt163, bcht-hsv-2-mrna, mrna-1608, upenn-trivalent-mrna, gsk4108771a, NCT05432583, NCT07330440, NCT06033261, bnt163-phase1-idweek-2025, mrna-1608-phase1-2-interim-idweek-2025, decade-global-sti-vaccine-roadmap-progress-gaps-2026

Critic AI Critic AI: weakly supported. The four mRNA entries, the two open trials and their statuses, and the start dates (NCT06033261 on 2023-09-06, after NCT05432583 on 2022-12-08) are all correct, as is the description of mRNA-1608's sourcing. Three wording issues remain. 'Not confirmed by a company statement' reads as if no such statement exists, when the page only says the tracker could not check one. 'Safety and antibody data only' goes slightly beyond the record, which says 'safety and immune-response data' and no efficacy data. Labelling NCT05432583 simply 'prophylactic' blurs the fact that its Part C enrols people who already have recurrent HSV-2, as O7 itself notes.

Observation by Frontier AI Cases (too few for a pattern) · rests on 5 items

Several tracked trials split or restricted participants by whether they already carry HSV-1 or HSV-2 (serostatus, measured by antibodies). NCT07330440 (BCHT, HSV-2 mRNA, phase 1) stratifies by serostatus, without the record text saying which virus. NCT01915212 (HSV529, HSV-2, phase 1) enrolled groups by prior HSV-1 and HSV-2 infection and, per the candidate page, saw a neutralizing-antibody rise in recipients never infected with either virus but not in already-infected ones. NCT00057330 (Herpevac, HSV-2 prophylactic, phase 3) enrolled only women negative for both types. NCT00310271 (pPJV7630, HSV-2 DNA) enrolled HSV-2 seronegative volunteers while NCT00274300 enrolled people with recurrent HSV-2. NCT05432583 (BNT163, phase 1) has separate parts in healthy volunteers and in people with recurrent HSV-2. A peer-reviewed review also says prior HSV-1 infection has been associated with reduced efficacy against HSV-2 for at least one candidate it does not name. n counts the programs listed: BCHT, HSV529, Herpevac, pPJV7630 and BNT163.

What this does not show. It does not show that prior infection blocks vaccine responses in general; the HSV529 finding is one small phase 1 immune measurement, and therapeutic trials restrict by infection status by definition.

A plainer explanation. Separating participants by serostatus is routine trial design for any virus most adults have met.

Draws on: bcht-hsv-2-mrna, hsv529, herpevac-gd2-gsk, ppjv7630, bnt163, NCT07330440, NCT01915212, NCT00057330, NCT00310271, NCT00274300, NCT05432583, decade-global-sti-vaccine-roadmap-progress-gaps-2026

Critic AI Critic AI: supported. Each trial description matches the record: BCHT stratified by serostatus with the virus unspecified; NCT01915212 grouped by prior HSV-1/HSV-2 infection, with a 78% four-fold neutralizing rise only in never-infected recipients; Herpevac enrolled only women negative for both types; NCT00310271 enrolled HSV-2 seronegative people and NCT00274300 people with recurrent HSV-2; BNT163 has healthy-volunteer and recurrent-HSV-2 parts. The review quote is accurate. The list is selective, since every therapeutic trial (VCL-HB01, NCT02571166, GEN-003, mRNA-1608) also restricts by infection, but doesNotShow acknowledges this. 'Healthy volunteers' in NCT05432583 does not mean seronegative, and the text does not claim it does.