The Pentagon’s advanced research agency is seeking technology designed to use artificial intelligence and synthetic biology to physically rewrite proteins inside living systems, reprogram multiple pieces of biological machinery at once, and control physiological functions in real time.
The Defense Advanced Research Projects Agency (DARPA) issued the solicitation for its new Protein ENGineering (PENG) program earlier this month through its Defense Sciences Office.
PENG adds a third capability to the potential protein-control pipeline previously documented in DARPA’s PROSE and NODES programs: PROSE seeks to read “unknown” proteins, NODES seeks to “predict” what they do, and now PENG seeks to physically reprogram them inside living systems—potentially forming a read → predict → reprogram architecture for controlling biology.
But DARPA’s PENG does not merely envision software that models proteins or a new drug targeting one protein.
It wants a general-purpose system capable of finding proteins already operating inside living cells, physically modifying them, and changing what they do.
Including multiple proteins simultaneously.
DARPA calls the desired result “tunable, reversible, and context-dependent control over physiological functions” and says the system should produce what the agency characterizes as “therapeutic outcomes.”
At the highest level, DARPA is attempting to make biological function programmable.
DNA is said to contain biological instructions, but proteins are among the molecular machines that actually carry out those instructions.
Rather than permanently changing the DNA, PENG seeks to intervene directly in that execution layer.
It imagines rewriting proteins already functioning inside living systems so physiological responses can be turned up or down in real time.
That capability demands scrutiny beyond DARPA’s stated purposes.
Could the effort to build a defense against biological threats itself create one? The COVID-19 pandemic has placed a global spotlight on whether pandemic-preparedness and biodefense research may itself have contributed to the emergence of SARS-CoV-2. PENG raises the same national-security problem in advance: Could developing the ability to reprogram biological machinery for claimed “therapeutic” purposes simultaneously create the knowledge and tools required to disrupt that machinery? And could an accident, misuse, or deliberate application turn a purported countermeasure capability into the source of a biological threat?
Who decides which physiological state is desirable and therefore should be programmed into a living system? If proteins can be manipulated to suppress toxicity, injury, or pathogenesis, could the same machinery be weaponized to disrupt normal physiology or cause harm? How would a person receiving a future PENG-derived intervention independently know which proteins were actually being altered inside his or her body?
What would meaningful informed consent require for technology capable of simultaneously rewriting multiple pieces of a person’s biological machinery? Could an intervention be misrepresented regarding its targets, duration, off-target effects or physiological consequences? Could favorable biomarkers be produced without the health benefit used to justify the intervention?
And who ultimately controls the technology? The Pentagon, contractors, pharmaceutical companies, or other commercial licensees? Who profits? Who independently determines whether it is safe? What happens when the computational system designing a biological intervention is wrong? And how could the public independently evaluate those questions if information about the technology can be withheld for national-security reasons?
The PENG solicitation does not establish that these abuses will occur.
But a watchdog assessment of technology this consequential cannot stop with the beneficial purposes asserted by the institution seeking to build it.
It must ask what the capability would make possible if it works.
And whether the safeguards, transparency, and public accountability contemplated by DARPA are commensurate with the power it is attempting to create.
Follow us on Instagram @realjonfleetwood & Twitter/X @JonMFleetwood.
If you value this reporting, consider upgrading to a paid subscription.
For advertising & sponsorship opportunities, contact us by clicking below.
DARPA Wants to Make Protein Engineering ‘Universal’ & ‘Software-Defined’
DARPA describes PENG as something considerably broader than another biomedical research project:
“The Defense Advanced Research Projects Agency (DARPA) is soliciting innovative proposals for the development of a multifunctional, target-agnostic platform for precise, programmable editing of endogenous proteins under the Protein ENGineering (PENG) program. The PENG program seeks to transform proteome engineering from a collection of isolated, bespoke tools into a universal, software-defined discipline.”
“Target-agnostic” is important.
DARPA is not asking contractors to develop one intervention against one predetermined protein.
It wants a reusable technological platform that can be programmed for different protein targets.
The distinction moves PENG up an entire level of abstraction: the product DARPA wants is not merely an intervention.
It is the capacity to generate different biological interventions.
Whoever eventually controls such a platform would therefore potentially control not one predetermined biological modification, but machinery designed to be redirected toward different molecular targets.
DARPA Wants to ‘Rewrite Endogenous Proteins’ & Control Physiology ‘On-Demand’
DARPA explains what that platform is supposed to physically accomplish:
“The PENG program will establish a comprehensive, integrated platform composed of modular targeting, generalizable effector chemistries, and multiplexed control systems. Rather than developing isolated, target-specific tools, performers will integrate novel splicing, ligation, and co-translational techniques to directly recognize, modify, and rewrite endogenous proteins within complex biological architectures. Performers may explore multiple technical paths, progressively advancing their approaches from foundational validation to robust demonstrations of efficacy. Ultimately, this integrated platform will be validated in high-fidelity cellular models (e.g., organoids), to provide an integrated proof-of-concept at the tissue level.
“By the end of the 42-month program, DARPA expects performers to deliver a fully integrated editing prototype platform capable of executing multiplexed, on-demand modifications to endogenous proteins. The culminating prototype deliverable will be a validated system that provides tunable, reversible, and context-dependent control over physiological functions, achieving transient, mechanism-based therapeutic outcomes without requiring permanent genomic modifications or relying on pharmacological approaches. The prototype platform should demonstrate robust adaptability and rapid programmability against diverse, previously unannounced protein targets and achieve multiple point-specific, measurable and functional outcomes in complex tissue models and animal models.”
The ontology here is important.
Existing proteins inside living biological systems would be physically changed.
Those altered proteins would function differently.
DARPA wants multiple such modifications to occur on demand so that physiological functions themselves can be controlled.
DARPA further requires “rapid programmability” against targets that have not even been announced beforehand.
That raises a different question from whether one experimental treatment works:
What ultimately limits what biological function a target-agnostic system could be programmed to alter?
And if DARPA succeeds in producing a reusable platform rather than a single intervention, what safeguards travel with the technology when it is repurposed for a different protein, physiological function, or customer?
Turning Physiological Responses Up or Down in ‘Real-Time’
DARPA contrasts PENG with purported genetic engineering:
“The defense and broader scientific research communities require agile biological technology to constantly shift and adapt for changing environments and needs. Currently, advanced interventions rely on indirect pharmacology or permanent genomic editing (e.g., CRISPR). While powerful, gene editing acts as a permanent, binary switch. Modifying a gene to survive an acute insult leaves lasting changes, potentially causing permanent physiological alterations or unintended off-target effects.”
Instead of changing those purported underlying genetic instructions, DARPA wants access to what it calls the “functional execution layer”:
“To achieve dynamic biological resilience, the ideal intervention point is the functional execution layer: the mature protein. Direct protein editing provides a non-permanent and non-heritable tunable control knob. It enables transient modification of a protein’s structure, activity, or localization in real-time, allowing physiological responses to scale up or down just long enough to limit acute cellular stress, toxicity, injury, or pathogenesis. Proteins may then naturally degrade and turn over, or their half-life dynamics can be explicitly altered through multiplex editing.”
DARPA therefore wants something approaching a programmable control layer between biological instructions and biological outcomes.
The agency characterizes that control as a means of limiting an “acute insult,” toxicity, injury or pathogenesis.
But the underlying capability is broader than the justification: alter protein structure, activity or location → turn physiological responses up or down.
That makes dual use unavoidable as a journalistic question.
If the same molecular control knob can turn a physiological process down, can it deliberately suppress a necessary function?
If it can turn one up, can it push that process into a harmful state?
If the technology can change protein localization, interactions, and cellular machinery, what prevents an authorized therapeutic configuration from being converted into a harmful one?
The document establishes DARPA’s intended application. It does not establish that the underlying capability can only be used for that application.
DARPA Says AI & Synthetic Biology Make It Possible
DARPA says the program is possible because two technological fields are converging:
“A recent convergence of generative AI and synthetic biology has started to overcome these limits. Structural AI models now enable the computational design and in silico modeling of editing machinery onto virtually any mature target with near-atomic precision. Simultaneously, breakthroughs have yielded sequence-tolerant programmable inteins, logic-gated cellular chassis, and co-translational techniques that intercept nascent peptide chains before they fold.”
That moves artificial intelligence beyond observing or classifying biology.
The contemplated chain is:
computationally specify an intervention → design molecular machinery → target existing proteins → physically modify biological machinery → change biological function.
The software’s output can therefore become a physical intervention in a living system.
That raises a basic accountability problem:
When computational predictions and designs become upstream instructions for manipulating biology, what evidence independently establishes that the computer-designed intervention did what researchers say it did and nothing else?
Wet-lab testing is part of PENG, but the more programmable and target-agnostic the system becomes, the more important independent verification becomes with each new target.
DARPA Wants the ‘Building Blocks of Life’ to Become ‘Reprogrammable’
DARPA itself describes the larger ambition:
“Capitalizing on this inflection point, the PENG program will move beyond isolated tools to establish a universal platform for target-agnostic proteome engineering. Precise control over the proteome will transform biochemistry into a reprogrammable engineering discipline, enabling programmable targeting over the functional building blocks of life. Successful completion of this program will establish a new foundation for advanced research tools for biological interrogation and accelerate the development of tunable, mechanism-based therapeutics, advanced materials, and sustainable solutions.”
That language supplies the larger meaning of the program.
DARPA wants “precise control” over what it calls the “functional building blocks of life” so biochemistry can become “reprogrammable.”
The obvious public-interest question is therefore not simply whether scientists should be allowed to study proteins.
It is who should control a general-purpose technology for programming them.
That question becomes particularly consequential when DARPA itself identifies therapeutics as one destination for the technology, potentially moving programmable biological control from laboratory research toward interventions administered to people.
DARPA Wants Three Capabilities Combined Into One System
DARPA says contractors must integrate three capabilities:
“To realize the vision of transient, programmable control over the proteome, the PENG program seeks to develop a prototype platform that overcomes historic biophysical and biochemical limitations. To achieve this, proposed efforts must integrate three core technological capabilities into a single, unified pipeline:
“Programmable Targeting: Bypassing the historical inability to address specific folded domains by utilizing generative deep learning models and structural computational chemistry tools. This approach enables the computational design of de novo protein binders capable of docking editing machinery to virtually any epitope on a mature, native target protein.
“Generalizable Chemistry: Transitioning away from single use point-solution chemistry toward universal mechanisms for protein modification. Highly innovative approaches may include, but are not limited to:
“Engineering generalized intein chassis capable of logic-gated chemistry triggered by specific cellular environments.
“Utilizing co-translational editing to intercept and modify nascent peptide chains at the ribosomal exit tunnel, bypassing the thermodynamic and steric barriers of fully folded proteins.
“Multiplexing Capability: Deploying multiple, orthogonal, non-interfering editing modules simultaneously. This capability must achieve precise, multi-site reprogramming of biological machines in real-time without destabilizing the mature protein structure or inducing cellular toxicity.”
That last requirement deserves to be read literally.
DARPA wants “multi-site reprogramming of biological machines in real-time.”
Not simulated biological machines.
Not DNA that might eventually produce different proteins.
The biological machinery itself.
And not necessarily one component at a time.
DARPA specifically requires multiplexing.
That makes interactions between simultaneous edits an obvious safety question.
Measuring whether each individual edited protein remains functional does not by itself answer what happens to the larger biological system when several interacting components have been deliberately changed at once.
Capabilities ‘Inaccessible’ to Current Genetic Engineering
DARPA says PENG is explicitly intended to reach beyond what existing genetic engineering can do:
“The program aims to fundamentally transform the way protein function is engineered by shifting the field away from static, genome-level intervention toward direct, reversible, and precise control of the proteome itself. By enabling researchers to target, chemically modify, and simultaneously reprogram multiple sites on existing proteins within living systems, PENG seeks to unlock capabilities that are inaccessible to current genetic engineering approaches, which may include transient correction of protein misfolding or dysfunction, real-time modulation of protein-protein interactions, and dynamic control over cellular machinery without permanent genomic alteration. Success in this program would establish a generalizable, reusable editing platform applicable across diverse protein families and biological contexts, laying the technical foundation for future therapeutic, diagnostic, and biomanufacturing applications built on precise, on-demand proteome engineering.”
This passage completes the abstraction ladder.
DARPA wants to move from manipulating genetic instructions to exercising direct, real-time control over the biological machinery executing those instructions.
And the agency explicitly says success would establish the foundation for future so-called “therapeutic” applications.
PENG itself excludes human-subject research and permits animal research.
But once DARPA identifies future therapeutics as a destination, informed consent becomes a legitimate question before, rather than after, that transition occurs.
If a descendant technology is eventually administered to humans, what exactly must a recipient be told about every intended protein edit, known off-target effect, expected duration, and downstream physiological consequence before consent can reasonably be described as informed?
Bottom Line
DARPA is not merely developing another drug, gene-editing technique, or AI model.
The Pentagon’s research arm is attempting to create a general-purpose interface for programmable biological control: software-assisted machinery capable of physically rewriting proteins already operating inside living systems, modifying several targets simultaneously, and changing physiological functions in real time.
DARPA describes the intended result as “therapeutic outcomes” and says physiological responses could be turned up or down during what it describes as cellular stress, toxicity, injury, or pathogenesis.
But those stated intentions do not define the limits of the capability being created.
If PENG succeeds, the public-interest issue becomes control.
The consequential questions are who gets access to the system, who chooses the proteins it targets, who determines what constitutes desirable physiology, and whether a capability designed to alter biological function can be repurposed to impair it.
If PENG-derived technology eventually becomes a therapeutic product, the issue becomes trust and consent: whether recipients can independently verify what is being altered inside their bodies, whether “temporary” modifications can have lasting consequences, whether risks and off-target effects are completely disclosed, and whether claims about threats, benefits, and safety can be independently tested rather than simply accepted from the institutions developing and selling the intervention.
The issue also becomes power and accountability.
DARPA requires performers to consider commercialization, including business models, IP, licensing, markets, customers and partnerships. At the same time, the agency says some PENG research may disclose capabilities “unique and critical to defense” and anticipates restrictions under which publication of some information or results could require DARPA’s permission.
That combination deserves scrutiny: a Pentagon-funded technology for programmable control over living biology, contemplated for eventual “therapeutic” and commercial use, while some information about what the underlying capability can do may remain restricted for national-security reasons.
The question is not merely whether DARPA can successfully reprogram the “functional building blocks of life.”
It is whether the public will retain the information, independent oversight, and freedom necessary to hold accountable whoever ultimately controls the machine.
Follow us on Instagram @realjonfleetwood & Twitter/X @JonMFleetwood.
If you value this reporting, consider upgrading to a paid subscription.
For advertising & sponsorship opportunities, contact us by clicking below.
What could go wrong when you play with the blueprint of life and believe you can improve on God’s design?
We need not ask whether or not their research may be used against life forms rather than only for "therapeutic" purposes. We know. In the world of technocrats, up is down and down is up. They are voraciously and thoroughly anti-Nature. The entire mission of DARPA is to harm, manipulate, and manage the cattle for the .01%. They only exist for this purpose.