Speech Without a Speaker

I. The Therapeutic Loosening Bites Deepest at the Gate of Speech

The preceding paper, Two Failure Modes, closed by pointing toward a direction it had not yet examined. What it diagnosed there were two modern failure modes of sīla under therapeuticization—one that keeps only the settling comfort the precept brings while discarding the form of its observance, and one that clutches the precept's outward shape while drawing off the volition within. And it noted, at the end, that this therapeutic dissolution of the precept comes on most fiercely at the gate of speech-action: when the "speech" that the precept against false speech addresses is itself, in the age of synthetic media and large language models, massively diluted, generated by proxy, and made hard to tell true from false, the "keeping the feeling, dropping the form" loosening bites deepest just here.¹ That "direction the next paper will look into expressly" is the present paper.

A line must be drawn first. The preceding paper set up two failure modes; what this paper diagnoses, keeping the utterance, dropping the volition, is not the addition of a third disease beside those two. On the contrary, it is a special case of that paper's "keeping the feeling, dropping the form" type, brought down onto the gate of speech-action—the same loosening, showing its deepest bite-mark on the matter of speech. There, the loosening lay in the parting of the precept's "form" from the "heart" within it; here, it lies in the parting of "speech" from the "volition" that issued it. The latter is only the former's specialized face at the speech gate, not a third disease standing on equal footing with it. This subordination, kept throughout the paper, is a line not to be crossed.

A second line must be drawn, lest the paper slide into an essay on "whether AI is good or bad." At every step, this paper asks one thing only: where, at this moment, is that intent to deceive?—never "is this technology a benefit or a harm." The technical particulars of synthetic media and language models, the gains and losses of policy, the menace of disinformation—none of these is the paper's subject. They are at most a scene, the place where the prying-open of the volition–speech joint first becomes visible at scale. The structure of the precept against false speech—that is, the one intent to deceive—is the subject from first to last; AI is only the place where the fracture in that structure can be seen with unusual clarity.

II. The Load-Bearing Condition of False Speech: The Intent to Deceive

To see this fracture clearly, one must first set the conditions of false speech upright. The Mahāprajñāpāramitā-śāstra, treating false speech, has a sentence that gathers all four conditions into one:

"False speech is: with an impure mind, wishing to deceive another, concealing the real, one utters words at variance—a verbal act arises—this is called false speech... It arises from the hearer's comprehension of the sound of the words; if there is no comprehension, then though the words be untrue, there is no offense of false speech."²

In this one sentence the four conditions stand complete. "Utters words at variance" is the words running contrary to what is known; "wishing to deceive another" is the intent to deceive; "a verbal act arises" is the utterance constituting karma; "arises from the hearer's comprehension of the sound" is the hearer having comprehended the words. Of the four, the last turn is the most telling: even untrue words spoken aloud, if the hearer never comprehends them at all, yield "no offense of false speech." So even the hearer's comprehension can keep false speech from being constituted; and the one condition that truly makes false speech a definite offense, that constitutes its moral substance, is the one stated first—"wishing to deceive another"—that one thought of deceiving another, which is the intent to deceive.

This gauge has a corollary often overlooked: the evil of false speech does not hang on the truth or falsity of the proposition spoken, but on that one thought of deceiving another. A person may speak words literally true while every word harbors the intent to deceive, meaning to lead the hearer into a false understanding—and by the gauge of false speech he stands not far from transgression; conversely, a statement objectively false, if wholly without the intent to deceive—a faulty memory, a slip of the tongue, a report of another's words sincerely believed true—the treatises plainly judge to carry "no offense of false speech." The scale of the precept weighs that one volition, not the truth-table of the sentence.

This is precisely the point at which the preceding paper's principle, it is volition that is karma, comes to rest at the gate of speech-action. The root of karma is volition; false speech is no more than that one thought of deceiving another, congealed into a sentence in the mouth. Everything said hereafter issues from this one line.

Two demarcations must be drawn in passing, to avoid overreach. First, this "gauge of volition" governs even the bodhisattva's case of permitted exception: the treatises allow a bodhisattva, in circumstances of benefiting beings, to speak untrue words "knowingly and with deliberation," judging that he is "of himself without a defiled mind" and so "commits no transgression."³ This paper takes from that only the principle of one gauge—whether it is a definite offense turns on whether that one thought is defiled or undefiled—and does not stray into the full account of how the bodhisattva precepts are undertaken, which is a subject for elsewhere. Second, within false speech there is the type called the great lie—claiming attainment of what one has not attained, falsely professing to have gained the dharma of a superior person—whose constitution of the offense and corresponding scriptural teaching have their own dedicated paper; this paper names it once and does not unfold it.⁴ What this paper keeps to is the face of false speech that runs through daily life: the gauge itself, that takes the intent to deceive as its load-bearing condition.

III. Speech Is the Disclosure of Volition: From Thought to Verbal Karma

If the root of karma is volition, then where, in false speech, does "speech" itself stand? The Abhidharmakośa, treating karma, first distinguishes "volition" from "the volitioned"—the stirring of the mind is the karma of volition; once volition has stirred and shows itself in body and speech, that is the volitioned karma—and "the body of the informative act of speech is just the sound of the speech": this verbal karma disclosed outward, though its body is that string of vocal sound, is "by the force of volition," produced by the deciding, motive volition that set it in motion.⁵ In other words, the sound of speech is not the source of the karma but its informing—the outward-shown face of that one volition.

Put plainly: a false utterance presupposes a speaker who has volition. The sound is only the outward showing of that one thought of deceiving another; volition comes first, speech after; speech is the disclosure by which that volition can be seen, heard, and comprehended by others. When this volume treated the binding undertaking and the precept-essence earlier, it ran the same line of reasoning: a precept can stand up in a person and be upheld only by the receiving mind at the moment of undertaking and the precept-essence thereby established and continuously guarded; the precept was never a set of external formal clauses laid over the body, but the establishing and continuance of one volition.⁶ So with verbal karma: whether a sentence is false speech or truthful speech is decided not in the string of sound but in the volition that issued it—volition sincere, the words are truthful speech; volition deceiving, the words are false speech.

This principle—that speech is the disclosure of volition—usually seems hardly worth stating: the one who speaks is, of course, the one who had a mind to speak; the truth or deceit in the words hangs, of course, on the speaker's volition. Just because this fit is so natural, we almost never think of "speech" and "the volition that issued it" as two things that could be parted. What the next section examines is precisely this natural fit, pried open for the first time by a new instrument for making speech.

IV. The AI Decoupling: Keeping the Utterance, Dropping the Volition

In the past, the joint between speech and volition was flush—where there was speech there was a volition that issued it, and to pursue the intent to deceive in a sentence one could always trace back to the one mind of the person who uttered it. Large language models and synthetic media, for the first time, pry this joint open at scale, prying out two ends that ought not to have parted.

One end is speech without volition. A machine can pour out fluent, human-seeming speech, ample enough to be comprehended, and behind that string of sound there is no thought of deceiving another—a machine has no mind that can be defiled, and so has no intent to deceive. By the gauge above, untrue words produced by a machine do not constitute false speech: to pin the offense of false speech onto an instrument that has no volition is a category error, just as one cannot judge a blade's sharpness to be the intent to kill. False speech still requires a speaker who has volition; the machine is not that speaker.

The other end is volition that utters no speech. That intent to deceive has not thereby vanished; it has migrated onto the person who uses this speech-making instrument—the one who gives the command, the one who deploys it, the one who borrows its mouth. He may harbor the thought of deceiving another without uttering the sentence himself: the sentence is written, generated, and issued for him by the machine. So the intent to deceive remains, yet no longer adheres to the heard string of sound; the volition that issued the words and the words that are comprehended are, for the first time, split into two places.

This paper names the phenomenon that joins these two ends keeping the utterance, dropping the volition (again: a diagnostic term this paper coins, not from scripture): the utterance remains—and grows more abundant, more lifelike, harder to tell true from false—while the volition behind it, and the bearing of responsibility for that volition, is washed away or diffused. The most everyday disclaimer—"I didn't say that, the AI wrote it"—is the spoken specimen of this decoupling: the sentence was indeed produced and published out of my intent to deceive, yet I have hidden that one volition behind an instrument with no volition. This is just where the preceding paper's "keeping the feeling, dropping the form" type bites deepest at the gate of speech-action—what loosens is no longer only the precept's form and heart, but "speech" and "the volition that issued it" themselves.

It is worth noting that several contemporary lines of thought all pass close by either side of this joint, and none lands squarely on the place where it fractures.

The first is the ethics of intention. Nicolas Bommarito, in Inner Virtue (2018), holds that the moral worth of an action hangs on the agent's inner intentions and the objects of his caring, not on the outward act itself—which is of a piece with "karma is volition" and "false speech is gauged by the intent to deceive," and this paper is glad to grant it.⁷ Yet his argument stops at the cultivation of inner character (how to become a person who cares about the right things), and does not single out for examination the concrete verbal-karmic joint of "how volition congeals into speech"—and the definite offense of false speech occurs precisely at that joint.

The second is the Yogācāra philosophy of language. Roy Tzohar, in A Yogācāra Buddhist Theory of Metaphor (2018), argues that all language is conventional designation (upacāra), that there is no simple one-to-one correspondence between word and referent, and that truth and falsity cannot be reduced to a direct match of the literal with the world.⁸ This deep observation loosens the naive understanding of "false speech = words that fail to match the facts," and this paper profits from it. Yet the gap he reveals lies on the side of reference—how a word points to a thing—whereas the gap in false speech lies on the side of intention—that it is the intent to deceive that drives the sentence. His philosophy of language, fine as it is, does not touch the joint where the moral and the mental meet.

The third is AI virtue ethics. Shannon Vallor, in Technology and the Virtues (2016) and the later The AI Mirror (2024), points out that technologically mediated communication erodes the various virtues on which honesty depends—as the channels of communication grow opaque and people grow estranged from the craft of expression, honesty itself is worn away.⁹ This is the line nearest to this paper's concern. Yet her remedy still falls on "the cultivation of virtue at the human end," and does not reach a certain category: the joint of "volition congealing into speech" has itself been outsourced to a system with no intent to deceive—the speech issued in place by a thing without volition, the volition held in secret in the heart of another person who utters nothing.

Set these three lines together and a seam appears that no one has filled: the ethics of intention sees the volition but does not isolate the verbal-karmic joint; the philosophy of language sees the word, but its gap is in reference and not intention; AI virtue ethics sees the erosion of technological mediation, but its remedy is still at the human end and does not reach the category of "the speech–volition joint being outsourced." No one reads false speech in the age of AI as a rupture of the volition–speech joint—the intent to deceive migrating off the heard string of sound. That seam is just what this paper means to identify.

V. A Natural Precept Cannot Be Dissolved: Diagnose, Don't Prescribe

Having identified the disease, must one then write a new prescription? Here a principle must be set, lest this diagnosis be misread as a piece of alarm: false speech is a natural precept; its evil lies in its nature, and is not something a change of speech-making instrument can dissolve.

The treatises distinguish natural precepts from proscribed precepts. Killing, stealing, sexual misconduct, and false speech—these four are precepts that "depart from natural offenses"; their evil lies in the nature of the matter, not made evil only because the Buddha prescribed against it; even were no such precept established, the very act of deceiving and harming another would remain evil.¹⁰ (Drinking and the like are proscribed precepts: the evil lies not in the nature of the matter but in its power to give rise to faults, and so the Buddha proscribed it.) False speech being one of the natural precepts, its moral substance hangs on the nature of that one intent to deceive—and the intent to deceive is a matter of the mind.

From this follows a crucial inference: what AI can change is the medium of speech; what it cannot dissolve is the moral substance of false speech. To hand a deceiving sentence to a machine to produce and to issue does not make the intent to deceive behind it vanish into thin air, nor make "I work the deception of another by borrowing an instrument with no volition" cease to be deception. The medium can dilute speech, can generate speech by proxy, can wash the speech until its origin is hard to trace—but it cannot touch the one volition the natural precept weighs. The problem of "volition" cannot be reduced to the problem of "medium"; what the "keeping the utterance, dropping the volition" diagnosed in the last section loosens is our capacity to trace that one volition, not the nature of that volition's evil. The disease lies in whether we are still willing to turn back and own that one volition—not in whether the precept still counts.

Precisely because the disease lies in owning, this paper, as a diagnosis, marks only where it lies and sets up no new prescription—the medicine is already stocked, with no need of new compounding. First, the positive pole: truthful speech, true speech. Among the ten courses of wholesome action, what abstaining from false speech accomplishes is just the true speech of the Tathāgata;¹¹ this pole stands at the exact opposite of false speech—one thought sincere, issued as words, and it is truthful speech. Second, the architecture this volume has built throughout: volition, the binding undertaking, the precept-essence. Whether a sentence is false speech or truthful speech is weighed back at the volition that issued it; that volition is established by the receiving at the moment of undertaking and guarded by the continuance of the precept-essence. The medicine cabinet is right here, and the door has stood open all along. What this paper does is at most to lead the patient before the cabinet and point: what you need has been here the whole time.

VI. Conclusion: The Question Left to the Sealing Paper

To gather what this paper has diagnosed. The load-bearing condition of false speech is that one thought of deceiving another—the intent to deceive; false speech is just that one volition congealed into verbal karma in the mouth, the sound no more than a face that volition shows outward. Large language models and synthetic media, for the first time, pry the volition–speech joint open at scale: the machine has speech without volition, the person has volition without needing to utter his speech; so the utterance remains while the volition departs—keeping the utterance, dropping the volition. Yet false speech is a natural precept; its evil lies in the nature of that one intent to deceive, which no medium can dissolve; so what this decoupling loosens is our capacity to turn back and own that one volition, not the force of the precept itself. The disease lying in owning, the medicine is nowhere else—in truthful speech at the positive pole, in the volition, undertaking, and precept-essence this volume has already built.

One word must be added, to complete the preceding paper's intent of "both ends in a single view." Keeping the utterance, dropping the volition is one end of this loosened joint; it has a mirror-image other face—keeping the words, dropping the truthfulness: every word verifiable, every sentence conforming to some formal rule of "disclosure according to the facts," while the whole stretch of speech still quietly deceives. Watertight in form, the volition long gone from the truthful. This face and "keeping the utterance, dropping the volition" make a pair, just as the preceding paper's two failure modes were two faces of one body. This paper only marks it and does not unfold it—leaving it there, as a mirror in which to look.

As for how this principle comes down into daily life—when, at this moment, I issue a sentence by borrowing a speech-making instrument, how to ask of myself, sentence by sentence, is that one volition behind this sentence still in it? do I still bear responsibility for it?—this layer of self-discernment, together with the gathering-up of this volume's whole doctrine, the synthesis of the round precept through the three scriptures, and the transition from precept into concentration, is all left to the sealing paper to handle expressly; this paper does not say it in advance.¹²

What this paper can do ends here: to point out for others the joint that has been pried open, and to add one reminder—whoever's mouth or whoever's instrument the speech is issued from, what matters, always, is whether behind the uttered sentence there is still a volition that can be owned, and that one is willing to own.

---

---

正法眼藏，不立文字——The value of the Dharma lies in the Dharma itself, not in the identity of the one who expounds it.

Pointing at the Moon ｜ Six Practice Gates · SĪLA 戒 ｜ CC BY-NC-SA 4.0 Scriptural citations collated against the CBETA Electronic Tripiṭaka ｜ github.com/pointing-at-the-moon/six-practice-gates

CBETA collation: the Chinese canonical citations in this paper — the Mahāprajñāpāramitā-śāstra (Dazhidu lun, T1509《大智度論》Mahāprajñāpāramitā-upadeśa-śāstraT25, No. 1509Kumārajīva (Later Qin, 405 CE)View on CBETA online reader ↗, the four conditions of false speech), the Abhidharmakośa (T1558《阿毘達磨俱舍論》Abhidharmakośa-bhāṣyaT29, No. 1558View on CBETA online reader ↗, vocal intimation and the four natural offenses) with Paramārtha's Kośa commentary (T1559《阿毘達磨俱舍釋論》T29, No. 1559View on CBETA online reader ↗), the Yogācārabhūmi (T1579《瑜伽師地論》Yogācārabhūmi-śāstraT30, No. 1579Xuanzang (Tang, 646–648 CE)View on CBETA online reader ↗, the bodhisattva exception), the Daśakuśalakarmapatha-sūtra (T0600, the fruit of abstaining from false speech), the Dharmaguptaka Vinaya (T1428《四分律》T22, No. 1428View on CBETA online reader ↗) and the Mūlasarvāstivāda Vinaya (T1442《根本說一切有部毘奈耶》T23, No. 1442View on CBETA online reader ↗, the great-lie precept) — were collated against the CBETA Electronic Tripiṭaka (2024 revision).

Pointing at the Moon · The Six Practice Gates · Volume IV SĪLA · Paper 29 · Part IX #2 · June 2026

Footnotes

1. See the direction marked at the close of this volume's P28, Two Failure Modes: the therapeutic dissolution of the precept comes on most fiercely at the gate of speech-action—the "speech" the precept against false speech addresses being, in the age of synthetic media and large language models, massively diluted, generated by proxy, and made hard to tell true from false, the "keeping the feeling, dropping the form" loosening bites deepest just here. The present paper is written in continuation of that. ↩

2. Mahāprajñāpāramitā-śāstra (Dazhidu lun), fascicle 13, Taishō vol. 25, no. T1509《大智度論》Mahāprajñāpāramitā-upadeśa-śāstraT25, No. 1509Kumārajīva (Later Qin, 405 CE)View on CBETA online reader ↗, p0157a; trans. Kumārajīva. The four conditions stand complete within the one sentence: uttering words at variance (contrary to what is known), wishing to deceive another (the intent to deceive), a verbal act arising (the utterance constituting karma), and arising from the hearer's comprehension of the sound (the hearer's comprehension); the closing turn—"if there is no comprehension, then though the words be untrue, there is no offense of false speech"—makes clear that the constitution of false speech hangs on the full set of four conditions, while the intent to deceive is its load-bearing one. ↩

3. Yogācārabhūmi (Yujia shidi lun), fascicle 41, Taishō vol. 30, no. T1579《瑜伽師地論》Yogācārabhūmi-śāstraT30, No. 1579Xuanzang (Tang, 646–648 CE)View on CBETA online reader ↗, p0517c. This paper takes here only one principle of the gauge—that whether it is a definite offense turns on whether the mind is defiled or undefiled—as corroboration of the principle that "false speech is gauged by the intent to deceive," and does not stray into the full account of how the bodhisattva precepts are undertaken. ↩

4. The great lie (claiming attainment of what one has not attained, falsely professing to have gained the dharma of a superior person), its constitution of the offense and corresponding scriptural teaching, is treated in detail in this volume's dedicated paper; for the vinaya-structural loci see the Dharmaguptaka Vinaya, T22 no. T1428《四分律》T22, No. 1428View on CBETA online reader ↗, fascicle 2 (the fourth pārājika, "claiming for oneself the dharma of a superior person"), p0577b, and the Mūlasarvāstivāda Vinaya, T23 no. T1442《根本說一切有部毘奈耶》T23, No. 1442View on CBETA online reader ↗, fascicle 10 (the training-precept on falsely professing to have attained the dharma of a superior person). This paper names it once and does not unfold it. ↩

5. Abhidharmakośa (Jushe lun), fascicle 13, Taishō vol. 29, no. T1558《阿毘達磨俱舍論》Abhidharmakośa-bhāṣyaT29, No. 1558View on CBETA online reader ↗. The sentence "the body of the informative act of speech is just the sound of speech" is at p0068c; the distinction of "volition" and "the volitioned karma," and the account of "by the force of volition" and "produced by the deciding motive volition and the stirring motive volition," precede it (around p0067b). Trans. Xuanzang. ↩

6. The sense of the binding undertaking and the precept-essence is set out in detail in this volume's P15 and P16; the establishing of "it is volition that is karma" in P28. ↩

7. Nicolas Bommarito, Inner Virtue (Oxford University Press, 2018). His account, holding the moral worth of an action to hang on inner intention and the objects of caring, is of a piece with "karma is volition"; yet it stops at the cultivation of inner character and does not single out the verbal-karmic joint of "volition congealing into speech." (Note: the §IV survey of the three interlocutors is a paraphrase-level reconstruction from disciplinary memory, not verbatim quotation; citation is at the level of the works as wholes, and no page-level pinpoint is claimed.) ↩

8. Roy Tzohar, A Yogācāra Buddhist Theory of Metaphor (Oxford University Press, 2018). His argument that all language is conventional designation (upacāra), with no one-to-one correspondence between word and referent; yet the gap he reveals lies on the side of reference, not on the side of intention. ↩

9. Shannon Vallor, Technology and the Virtues (Oxford University Press, 2016); The AI Mirror (2024). Her point that technologically mediated communication erodes the virtues on which honesty depends; yet her remedy falls at the human end of virtue cultivation, and does not reach the category of "the volition–speech joint being outsourced to a system with no intent to deceive." ↩

10. "Departing from killing... down to departing from false speech, these four depart from natural offenses," see Abhidharmakośa, fascicle 14, Taishō vol. 29, no. T1558《阿毘達磨俱舍論》Abhidharmakośa-bhāṣyaT29, No. 1558View on CBETA online reader ↗, p0075b (trans. Xuanzang); for Paramārtha's rendering see the Abhidharmakośa-śāstra commentary (Apidamo jushe shilun), fascicle 11, T29 no. T1559《阿毘達磨俱舍釋論》T29, No. 1559View on CBETA online reader ↗, p0232c. Killing, stealing, sexual misconduct, and false speech are the four natural offenses, evil in the nature of the matter; drinking and the like are proscribed precepts. ↩

11. The virtue accomplished by abstaining from false speech—down to "when one later becomes a Buddha, one attains the true speech of the Tathāgata"—see the Daśakuśalakarmapatha-sūtra (Shi shan ye dao jing), fascicle 1, Taishō vol. 15, no. T0600, p0158a–b (the passage on abstaining from false speech begins at 0158a; the line "attains the true speech of the Tathāgata" is at 0158b). ↩

12. The daily-life self-discernment, the gathering-up of this volume's whole doctrine, the synthesis of the round precept through the three scriptures, and the handoff from precept to concentration, all belong to this volume's sealing paper (P30); this paper does not set up their account in advance. ↩