On 1 August 2026 Yonatan Chemla and Christopher A. Voigt published a report in Trends in Biotechnology, entitled “Responsible biotechnology: An analysis of the Asilomar 2025 entreaties”. This caught my attention, because, just over ten years ago, I wrote a blog post about the first Asilomar conference held in 1975 entitled ‘Ta(l)king responsibility’. And, of course, I also wondered what an ‘entreaty’ was.
Both conferences dealt with genetic engineering, from recombinant DNA in 1975 to AI genome synthesis in 2025. I’ll focus here not on the science but on the way that responsibility about science was assumed, staged and performed and how responsibility was conferred or deferred, using speech act theory as a theoretical lens. John Austin pointed out in 1955, in a book called How to Do Things with Words, that language doesn’t just describe things but carries out social actions. This came to be known as speech act theory.
Performing responsibility – performing responsibly?
In 2015, at the height of discussions about ‘responsible research and innovation’, I examined the way scientists at the original 1975 Asilomar conference didn’t so much accept oversight as perform it: standing up, in public, and narrating their own (“self-imposed”) restraint into existence before anyone could impose it on them from outside. Responsibility was in this case not something that was assumed, but a speech act carried out in front of witnesses. In ‘talking’ responsibility the scientists were also ‘taking’ responsibility.
Fifty years later, the “The Spirit of Asilomar and the Future of Biotechnology” conference was the sequel performance, including some of the old performers¹, but performed differently, in a different genre and by performing different speech acts. But, as we shall see, there were also continuities between the two conferences, especially about the relation between performers and audience or publics. Before I go into this, a quick look at the two conferences.
Asilomar 1975 and 2025
In 1974, molecular biologists began to express concerns about possible dangers posed by new methods they had developed to splice and recombine DNA. They therefore temporarily stopped work using this new technology. A year later, in 1975, they organised a conference in Asilomar, California, in which 140 to 150 people, mostly leading molecular biologists, along with a small group of lawyers, physicians, science journalists, and government officials, participated.
The conference focused on safety issues around an emergent threat: recombinant DNA and genetic engineering. At that conference “the biologists decided among themselves what restrictions should apply to various types of genetic manipulation” (Rasmussen, 2015). They were criticised by social scientists and historians of science for not bringing members of the public into core discussions about risk and responsibility and thus endangering trust in science. (For all the background gossip as well as in-depth ethical and historical discussions, see Matthew Cobb’s The Genetic Age!)
The conference ended with the publication of a set of voluntary safety guidelines. It lifted a self-imposed research freeze and replaced it with a risk-based tier system (minimal, low, moderate, and high containment) for genetic engineering, which later shaped official government policy.
In 2025, 300 participants from 25 countries and from many disciplines and domains of life met in Asilomar, not to commemorate the 1975 conference but to discuss new challenges and threats. There were “representatives from governments and private industry, as well as journalists, artists, social scientists”, ethicists and historians, including Matthew Cobb, for example, from whose personal reflection I just quoted. The conference also explicitly included early career researchers and indigenous organisations.
There was not only a difference in participation but also in thematic focus and output. As Cobb points out, rather than being organised around one high-stakes safety issue as in 1975, Asilomar 2025 split into five thematic groups: pathogens and bioweapons research, AI and biotech, synthetic cells and mirror life, environmental and non-conventional containment, and a more loosely defined group on “Framing Biotechnology’s Future.” The conference ended with the publication of 27 so-called collaborative ‘entreaties’ “addressing modern synthetic biology, AI integration, and biosecurity” – topics that loom ever more largely in an AI where AI automation and weaponisation amplify each other and topics that were blended out in 1975.
Shifts in genres and speech acts
The original Asilomar conference produced one document: a Summary (Consensus) Statement, deliberated collectively and issued as a single voice. In speech act terms, this was first and foremost a commissive: the scientists bound or committed themselves to a future course of action, which was not to proceed without physical and biological containment. Commissives are one of the speech act categories that the philosopher of language John Searle identified in his refinement of a taxonomy first set out by J. L. Austin in the 1960s: unlike a mere statement of fact, a commissive obligates the speaker to something.
That self-binding commissive was taken up by the US National Institutes of Health in 1976 and converted into official guidelines. One can call this second act a declaration in stricter speech act terms, that is to say, an institutional utterance that changes reality by virtue of being said, by someone with the authority to say it. Take for example ‘I hereby declare you man and wife’, or ‘the meeting is adjourned’. Here the speech act is self-executing, saying it is doing it. A commissive made by an unauthorised group of scientists became, through the NIH’s uptake, something closer to a declaration: “we hereby restrict recombinant DNA to these biosafety tiers”, binding not just on the scientists who said it but on the whole field.
In contrast with the 1975 conference, the 2025 Asilomar conference produced twenty-seven documents. Instead of one document we have a whole anthology of documents and a shift in genre. The documents were called “Entreaties”, with an ‘entreaty’ defined on a repository of these documents as “a humble but urgent request for engagement”. In terms of speech act theory, an entreaty belongs to what Searle called the category of directives: attempts by a speaker to get the listener to take a specific action, not unlike demands or requests; more a petition than a promise.
This is where the two conferences diverge beyond just genre. A commissive is the kind of speech act that a regulator can take up and ratify into a declaration, because there is something that can be converted. This is what the NIH did in 1976. A directive has no such content: nobody who signs an entreaty has bound themselves to anything, so there is nothing equivalent for the NIH, the WHO, or any national regulator to convert and ratify into policy fifty years on.
Each Asilomar entreaty was written by a subset of attendees who were not necessarily in conversation with each other; each was put to a vote; and each needed only 10% support to be endorsed and published. The 2025 Asilomar conference did not speak with one voice and didn’t come up with one commissive. It spoke in a chorus of multiple voices.
In their synthesis document, Chemla and Voigt organised the 27 entreaties into six thematic groups: regulatory, governance, release, data, research, community and they work hard to reconcile positions that don’t obviously agree with each other. Instead of forcing a consensus, dissensus is accepted. That is to be applauded.
Shifts in responsibility
The shift from statement to anthology, from commissive to directive, and from one voice to a plurality of voices signals a shift in the kind of responsibility being performed. A single consensus statement claims to speak for a field. It says: this is what we, collectively, will restrain ourselves from doing in terms of genetic engineering. An anthology of individually (co-)authored amd endorsed entreaties instead performs pluralism. It says: look how many of us are worried, from how many angles, in how many registers or genres (there were statements, proposals, frameworks, videos, AI simulated debates and more).
This polyphony of voices comes across as more honest and responsible than a single voice claiming a consensus that might never have been there. But it also does something else: it disperses accountability and defers responsibility. Twenty-seven entreaties, none of them binding, none of them requiring the others’ agreement, add up to a chorus rather than a single ratifiable commissive; and a chorus is much harder to hold to account.
A consensus statement performs (or claims to be) a commissive: a group binding itself to a shared position, and, as we have seen, binding itself in a form a regulator could later ratify into something more binding still. An entreaty, by contrast, is a directive dressed as a petition (“a humble but urgent request for engagement”) addressed outward, to someone else, asking them to act, with no commitment of the entreater’s own for anyone to ratify.
This is quite a difference, indeed a weakening, in what speech act theorists call illocutionary force (the intended function or purpose of words used) between 1975 and 2025. The 27 entreaties look like responsible innovation, but in effect, nobody who signs an entreaty has committed themselves to do anything. The entreaties ask the field, or regulators, or “future stakeholders,” to engage. This is responsible innovation without commitment or obligation. It doesn’t confer responsibility; it defers it. It’s performance but not performative.
Public performance, but what about the public?
Speech acts can, under the right circumstances, perform real actions in the world and have real consequences. For example a promise made in public is harder to break than one made in private. The speech acts made during the 1975 and the 2025 Asilomar conferences were performed in public. In fact, in the case of the 2025 one in front of a diverse public of 300 witnesses and now a journal readership. But there was a difference.
In 1975, a group of scientists publicly committed themselves to restrain and to refrain from certain actions from then on. This produced the biosafety-level infrastructure we still use today. Something got built, not just performed.
In 2025, a diverse group of people, with scientists at the centre, publicly ‘entreated’ others to refrain from certain actions in front of a diverse public. This act did not confer responsibility on anyone in particular; instead, responsibility was deferred pending public engagement.
In both cases the position of members of the public in this performance or process was precarious. In 1975 engagement with them was largely non-existent, as scientists took responsibility largely on their own initiative; in 2025 public engagement was achieved, but it was also used to defer responsible action into the future.
There is also a more deeply engrained and recurring theme emerging from this brief comparison of two conferences. Each time genetics and genomics make significant advances, be it recombinant DNA in 1975 or AI-designed genomes in 2025/26, it seems that the public performance of responsibility is scientists narrating their own responsibility into being, in scientists’ own chosen genre and on their terms. The public appears mainly as an audience for the performance or a stakeholder to be engaged with or appealed to afterwards, rarely as a co-author of what responsibility should mean or who gets to withhold consent. Fifty years apart, the entreaties and the commentary are still asking us to trust the performance and the performers rather than co-produce the script.
Footnote:
Matthew Cobb, personal communication, August 2026: among those present at the 2025 roundtable ‘Voices from 1975’ were Stan Cohen, Janet Mertz and Michael Rogers.
Image: Watercolor by James G. Swan depicting the Klallam people of chief Chetzemoka at Port Townsend, with one of Chetzemoka’s wives distributing potlatch
Leave a Reply