diff --git a/src/vs/code/node/sharedProcess/contrib/voiceTranscriber.ts b/src/vs/code/node/sharedProcess/contrib/voiceTranscriber.ts
index d9a9979dcb3..210570ef952 100644
--- a/src/vs/code/node/sharedProcess/contrib/voiceTranscriber.ts
+++ b/src/vs/code/node/sharedProcess/contrib/voiceTranscriber.ts
@@ -32,6 +32,8 @@ export class VoiceTranscriptionManager extends Disposable {
 
 class VoiceTranscriber extends Disposable {
 
+	private static MAX_DATA_LENGTH = 30 /* seconds */ * 16000 /* sampling rate */ * 16 /* bith depth */ * 1 /* channels */ / 8;
+
 	private readonly transcriptionSequentializer = new TaskSequentializer();
 
 	private requests = 0;
@@ -81,7 +83,14 @@ class VoiceTranscriber extends Disposable {
 			}
 		}
 
-		this.data = this.joinFloat32Arrays(this.data ? [this.data, ...newData] : newData);
+		const dataCandidate = this.joinFloat32Arrays(this.data ? [this.data, ...newData] : newData);
+
+		if (dataCandidate.length > VoiceTranscriber.MAX_DATA_LENGTH) {
+			this.logService.warn(`[voice] transcriber: refusing to accept more than 30s of audio data`);
+			return;
+		}
+
+		this.data = dataCandidate;
 		const data = this.data.slice(0);
 
 		this.requests++;
diff --git a/src/vs/platform/voiceRecognition/node/voiceRecognitionService.ts b/src/vs/platform/voiceRecognition/node/voiceRecognitionService.ts
index ca569eae27e..a2a1d3218e9 100644
--- a/src/vs/platform/voiceRecognition/node/voiceRecognitionService.ts
+++ b/src/vs/platform/voiceRecognition/node/voiceRecognitionService.ts
@@ -44,11 +44,11 @@ export class VoiceRecognitionService implements IVoiceRecognitionService {
 
 		const now = Date.now();
 
-		const voiceModule: { transcribe: (audioBuffer: { channelCount: 1; sampleRate: 16000; sampleSize: 16; channelData: Float32Array }, options: { language: string | 'auto'; suppressNonSpeechTokens: boolean }) => Promise<string> } = require.__$__nodeRequire(modulePath);
+		const voiceModule: { transcribe: (audioBuffer: { channelCount: 1; samplingRate: 16000; bitDepth: 16; channelData: Float32Array }, options: { language: string | 'auto'; suppressNonSpeechTokens: boolean }) => Promise<string> } = require.__$__nodeRequire(modulePath);
 
 		const text = await voiceModule.transcribe({
-			sampleRate: 16000,
-			sampleSize: 16,
+			samplingRate: 16000,
+			bitDepth: 16,
 			channelCount: 1,
 			channelData
 		}, {
diff --git a/src/vs/workbench/services/voiceRecognition/electron-sandbox/workbenchVoiceRecognitionService.ts b/src/vs/workbench/services/voiceRecognition/electron-sandbox/workbenchVoiceRecognitionService.ts
index 17260848ccf..e2e94971c3a 100644
--- a/src/vs/workbench/services/voiceRecognition/electron-sandbox/workbenchVoiceRecognitionService.ts
+++ b/src/vs/workbench/services/voiceRecognition/electron-sandbox/workbenchVoiceRecognitionService.ts
@@ -67,8 +67,8 @@ export class WorkbenchVoiceRecognitionService implements IWorkbenchVoiceRecognit
 
 	declare readonly _serviceBrand: undefined;
 
-	private static readonly AUDIO_SAMPLE_RATE = 16000;
-	private static readonly AUDIO_SAMPLE_SIZE = 16;
+	private static readonly AUDIO_SAMPLING_RATE = 16000;
+	private static readonly AUDIO_BIT_DEPTH = 16;
 	private static readonly AUDIO_CHANNELS = 1;
 
 	constructor(
@@ -96,8 +96,8 @@ export class WorkbenchVoiceRecognitionService implements IWorkbenchVoiceRecognit
 
 			const microphoneDevice = await navigator.mediaDevices.getUserMedia({
 				audio: {
-					sampleRate: WorkbenchVoiceRecognitionService.AUDIO_SAMPLE_RATE,
-					sampleSize: WorkbenchVoiceRecognitionService.AUDIO_SAMPLE_SIZE,
+					sampleRate: WorkbenchVoiceRecognitionService.AUDIO_SAMPLING_RATE,
+					sampleSize: WorkbenchVoiceRecognitionService.AUDIO_BIT_DEPTH,
 					channelCount: WorkbenchVoiceRecognitionService.AUDIO_CHANNELS,
 					autoGainControl: true,
 					noiseSuppression: true
@@ -109,7 +109,7 @@ export class WorkbenchVoiceRecognitionService implements IWorkbenchVoiceRecognit
 			}
 
 			const audioContext = new AudioContext({
-				sampleRate: WorkbenchVoiceRecognitionService.AUDIO_SAMPLE_RATE,
+				sampleRate: WorkbenchVoiceRecognitionService.AUDIO_SAMPLING_RATE,
 				latencyHint: 'interactive'
 			});