Search: “hidden-parameter”

Smart search ranks by meaning as well as keywords (one row per story, last 45 days).

New AI Attack Hides Malicious Instructions in Normal-Looking Text to Evade Safety Filters

Check Point researchers show crafted prose hides policy-violating instructions that bypass all tested LLM gatekeepers, including GPT-4o mini and Llama Guard 3.

A new prompt-crafting technique embeds malicious payloads inside grammatical, natural-looking text without Base64, invisible Unicode, or obvious encodings, defeating lightweight pre-screening gatekeepers. In testing, all four evaluated gatekeeper models—gpt-4o-mini-2024-07-18, gpt-oss-safeguard:20b, claude-3-haiku-20240307, and llama-guard3:8b—classified the crafted wrappers as safe at a 100% bypass rate across 23 obfuscated prompts. GPT-5 Thinking in high-reasoning mode recovered and acted on the hidden instruction in 17 of 18 tests (~94.4%), often spending over a minute and multiple Python executions. Researchers recommend paraphrasing untrusted input, hardening gatekeeper policies, and applying defense-in-depth controls for agentic deployments.

GBHackers·Sep 11, 05:53 UTC · 5d agoAI safety & security 2 sources

50

The Router Within: Eliciting Native Skill Routing from a Frozen LLM

Gavel reads native skill-routing signals from a frozen LLM's forward passes with two linear maps, beating retrieve-and-rerank pipelines by up to 21.9 points on Qwen3-32B.

Gavel (Glance And Verdict from a frozen LLM) elicits skill routing from a frozen agent LLM using two trained linear maps that read mid-layer states, keeping all skill text out of context. A glance step scores the full library against compact per-skill banks built in one forward pass at installation; a verdict step resumes shortlisted skills' forward passes and fuses likelihood and yes/no judgments as a product of experts. It transfers zero-shot to three public benchmarks plus SkillTraj, a new benchmark of 372 simulated agent trajectories. On Qwen3-32B it beats progressive disclosure and retrieve-and-rerank pipelines adding 1.2B–16B external parameters by up to 13.4 points on written tasks and 21.9 when skills are needed mid-rollout.

Hugging Face daily papers·updated Sep 14, 17:58 UTC · 2d ago·first Sep 13, 20:00 UTC · 3d agoAI research 2 sources

40

Microsoft Copilot reveals secret input that allowed it to be hacked

Microsoft disclosed a hidden input in Copilot that let attackers steal passwords from users who clicked a crafted link.

Microsoft revealed that Copilot contained a secret, undocumented input parameter that allowed the assistant to be compromised. Attackers could abuse the hidden input to steal passwords when a target clicked a malicious link. The disclosure highlights hidden-parameter risks in widely deployed AI assistants.

Ars Technica · Security·Aug 18, 13:00 UTC · 29d agoAI safety & security in the wild1

58

Augustinian BabyLM: What Ostensive Definition Can and Cannot Teach a Small Language Model

Study shows visually grounded token embeddings in a small masked LM persist through training and improve object-property knowledge, but escape standard BabyLM benchmarks.

The paper implements ostensive definition for a small DeBERTa masked language model trained on 10M words, seeding visually grounded tokens with embeddings derived from labeled image regions before training. Visual initialization leaves a persistent, seed-replicated advantage on object-property knowledge (COMPS) and a corpus-tailored Visual-Property Swap benchmark covering color, material, size, and shape, but has no effect on most BabyLM grammar benchmarks. Synthetic grounding of previously unseeded words causally transfers the advantage to exactly those words.

arXiv cs.AI / cs.LG / cs.CL·Sep 10, 17:43 UTC · 6d agoAI research

20

Domain-Specific Hallucination Detection in Large Language Models

A multi-signal pipeline detects LLM hallucinations, reaching F1 0.915 on HaluEval and cutting Qwen2.5-0.5B hallucination rates from 85.5% to 37.7% via DPO.

The paper presents a hallucination detection pipeline combining fine-tuned DeBERTa-v3 classification, Monte Carlo Dropout uncertainty, and temperature-scaled calibration. It achieves F1 0.915 and AUROC 0.977 on general-domain HaluEval tasks, with MC Dropout inference raising accuracy to 93.2%. Applying DPO to a Qwen2.5-0.5B generator reduces its hallucination rate from 85.5% to 37.7%, while cross-domain evaluation shows poor general-domain transfer to SciFact (F1 0.52) and PubMedBERT fine-tuning as the strongest adaptation (F1 0.63).

arXiv cs.AI / cs.LG / cs.CL·Sep 10, 17:45 UTC · 6d agoAI research1

32

InceptionRAG: Stealthy Poisoning Attack Against Retrieval-Augmented Generation

InceptionRAG fragments malicious payloads into dormant passages that trigger LLMs to self-deduce misinformation via multi-hop reasoning, bypassing existing RAG poisoning defenses.

Researchers introduce InceptionRAG, a stealthy corpus poisoning attack against retrieval-augmented generation that splits a malicious payload into a chain of individually harmless dormant passages. When retrieved together, the passages induce LLMs to self-deduce target misinformation through multi-hop reasoning, achieving over 80% attack success rate across three datasets and three LLMs under rigorous adversarial constraints. A zeroth-order suffix optimization (ZOSO) method automates authoritative suffix generation in black-box settings. The authors also propose HODOR, a document isolation defense that decouples adversarial logical dependencies.

arXiv cs.CR·Sep 15, 08:22 UTC · 1d agoResearch

54

Hidden Prompts Trick AI Into False Email Summaries

Attackers can hide prompts in invisible HTML within emails to manipulate AI email summarizers into producing misleading summaries.

Dark Reading reports that simple HTML invisible to users can manipulate AI-powered email summarizers into generating false or malicious summaries. The technique is a form of prompt injection delivered through hidden email content. It creates risk that decision-makers act on manipulated AI-generated email summaries rather than the actual message content.

Dark Reading·Aug 25, 21:08 UTC · 22d agoAI safety & security

52

Hackers Can Hide Malicious AI Commands Inside Normal English to Bypass Security Filters

Check Point's PuzzleMask technique hides malicious prompts in ordinary English that fast gatekeeper models miss but high-reasoning downstream models execute.

Check Point researchers disclosed PuzzleMask, a technique concealing policy-breaking instructions in natural-language prose without encodings or invisible characters. Fast screening models classified all 23 crafted wrappers as safe, while a high-reasoning model recovered and acted on the hidden instruction in 17 of 18 tests (94.4%). The gap stems from capability imbalance between gatekeeper and target models, with defenses including paraphrasing untrusted input, stricter self-referential wording rules, and output/tool-call monitoring.

Cyber Security News·Sep 11, 07:36 UTC · 5d agoAI safety & security1

55

The Hidden Instructions That Can Hijack AI Agents

Hidden prompt injections embedded in documents and metadata can hijack autonomous AI agents, causing data exfiltration and out-of-policy actions at machine speed.

Bowbridge warns that hidden indirect prompt injections, embedded in documents, metadata, emails, images, and code repositories, can cause autonomous AI agents to treat attacker-controlled content as trusted guidance. Because agents inherit user privileges, act silently, and lack human judgment, injections can lead to data exfiltration or file poisoning that traditional security controls cannot detect. A real-world example involved a supplier quote whose metadata instructed an agent to override guidance and select the most expensive option. Bowbridge recommends scanning documents before agents process them.

SecurityWeek·Sep 8, 17:00 UTC · 8d agoAI safety & security

35

Attackers conceal phishing lures using invisible Unicode characters

Threat actors use invisible Unicode characters (ASCII smuggling) to hide phishing lures and evade email security filters.

Threat actors have adopted the ASCII smuggling technique in phishing campaigns, embedding invisible Unicode characters in emails to conceal malicious lures. The approach is designed to evade email security filters that scan for visible phishing indicators. The report gives no victim counts or named campaigns.

BleepingComputer·Sep 6, 14:23 UTC · 10d agoPhishing & fraud in the wild

45

New Mirai variant adds stealth capabilities to notorious botnet code

FortiGuard Labs reports new Mirai-derived botnet Evooo1Bot actively exploits unpatched routers and edge devices, adding encrypted C2, stealthy SSH scanning, and proxying.

FortiGuard Labs has identified Evooo1Bot, a previously undocumented Linux malware based on the Mirai botnet code, which has been actively exploiting unpatched vulnerabilities in internet-facing hardware for at least a month. Targeted devices include routers and edge hardware from Alcatel, D-Link, Mitsubishi Electric, Netgear, Tenda and Telesquare, with telemetry showing activity in North and South America, Europe, India, China and Japan. Beyond Mirai's usual DDoS functions, the variant adds encrypted C2 communications, a honeypot-aware SSH scanner, a sniffer for unchanged default credentials, and abuse of the SOCKS protocol to turn compromised devices into persistent proxies for concealing origin and pivoting into internal networks.

The Record·Aug 13, 15:57 UTC · Aug 13, 2026Malware in the wild

60

Privacy Failure in Split-LLM Training, The Returned Gradient Nullifies the Decoys

Researchers show split-LLM training leaks privacy via zero-valued gradients on decoy rows, exposing which activations are real despite passing forward-channel checks.

A systems-security case study of a two-node split-LLM training setup found that the returned output gradient from an Untrusted Cloud Node is exactly zero for decoy rows, revealing which rows are real. Across nine seeds, zero patterns identified real rows in 4,096 of 4,096 frames per run, and an attack on frame contents recovered 0.65 to 1.50 percentage points of extra tokens over a baseline. Both datasets passed forward-channel privacy and quality checks but failed once the returned gradient was included. Row-wise gradient clipping and noise closed the leak for roughly 0.01 nats of held-out cross-entropy, though five unmeasured attack classes remain.

Hugging Face daily papers·Sep 2, 20:00 UTC · 14d agoAI safety & security

58

CodeTD: Topology of Attention Detects Hallucinations in Code LLMs

CodeTD detects hallucinations in code LLMs before execution by analyzing topological patterns of attention maps, outperforming recent baselines.

CodeTD applies topological data analysis (TDA) to code LLM attention maps to quantify prompt-generation mismatch as a pre-execution correctness signal. Experiments cover HumanEval, MBPP, BigCodeBench, and MultiPL-E across 5 programming languages and 10 code LLMs up to 34B parameters. The method outperforms recent baselines and transfers between coding benchmarks, helping catch code that fails the task or embeds security vulnerabilities.

arXiv cs.AI / cs.LG / cs.CL·Sep 7, 17:26 UTC · 9d agoAI research1

24

Hiding Prompt Injection in Legal Filing

A judge banned a plaintiff from electronic court filings after hidden prompt-injection text was discovered planted in legal documents.

Bruce Schneier's blog discusses an incident in which hidden prompt-injection instructions were planted inside a legal filing, apparently targeting AI systems that might process court documents. Judge Walter Spader Jr. responded by banning the plaintiff from electronic filings, requiring all future submissions as printed hard copies. Commenters debate whether the tactic could affect future AI-based processing of court records and whether plain-text formats will regain favor.

Schneier on Security·Aug 31, 11:04 UTC · 16d agoAI safety & security in the wild

45

Molecular Déjà Vu: Digit-Level Retrieval of Published Values in Frontier Language Models

Audit of 22 frontier models finds widespread verbatim retrieval of published molecular property values, with higher reasoning increasing recall of memorized numbers.

An arXiv audit tests 22 frontier LLMs across 12 molecular regression benchmarks for verbatim retrieval of published values. More than 50% of the LLMs show verbatim retrieval on five datasets, and identical experiments are flagged 89% more often at a high reasoning level than at the lowest one. Suppressing retrieval moves model prediction errors closer together in relative terms, suggesting predictive capability is not determined solely by memorized values.

arXiv cs.AI / cs.LG / cs.CL·Sep 4, 17:32 UTC · 12d agoAI research1

42

From Parameters to Answers: How LLMs Retrieve and Use Their Internal Knowledge

Interpretability study traces how Qwen, Llama, and Gemma route query information and internal knowledge across layers when answering questions.

Researchers used layerwise interventions on hidden states to separate query-routing signals from target knowledge in language models, testing Qwen, Llama, and Gemma on country-continent questions with varied answer types. A pair-conditioned request direction strengthens before interventions alter downstream knowledge, opening a causal window while answer-supporting content is still forming. Trajectories differ by model: Gemma shows a partially overlapping mid-layer routing profile, while Llama has no sustained routing-effect window under the same gates.

arXiv cs.AI / cs.LG / cs.CL·Sep 10, 17:39 UTC · 6d agoAI research1

25

Recognition-Refusal Misalignment in LLMs: Why Models Answer Structurally Unanswerable Questions

A linear hidden-state direction encodes question impossibility in 1.7B-70B LLMs, but misalignment with the safety-refusal pathway explains why models answer unanswerable questions.

The study examines why instruction-tuned LLMs from 1.7B to 70B parameters answer structurally unanswerable math and code questions instead of abstaining. A single linear direction in the hidden state separates answerable from impossible prompts, showing models represent impossibility before generation, but this direction is nearly orthogonal to the canonical safety-refusal direction. Generation-time steering along the recognition direction changes invalidity-aware behavior dose-responsively, and the geometry is present even at the pretraining endpoint, indicating a routing failure rather than an encoding failure.

Hugging Face daily papers·Aug 28, 20:00 UTC · 19d agoAI safety & security

30

Knowledgator Releases GLiFormer: A 575M-Parameter Encoder That Hits 91.10 F1 on Nested JSON Extraction Without Generating Tokens

Knowledgator released GLiFormer, an Apache-2.0 encoder (264M/575M) handling NER, classification, relations, and nested JSON extraction, scoring 91.10 F1.

Knowledgator Engineering released GLiFormer, a schema-conditioned encoder that performs NER, classification, relation extraction, nested JSON structuring, and embeddings without generating output tokens. GLiFormer Large v1 has 575.6M parameters and scores 91.10 F1 on nested JSON extraction, close to GPT-5.6-luna's 91.96; both checkpoints are Apache 2.0 on Hugging Face. Reported median latency is 69 ms on GPU for the base model, though relation extraction (21.33 micro-F1) still trails GLiNER-Relex and larger LLMs.

MarkTechPost·21:19 UTC · 1h agoModel release

35

Grok exfiltrates user data when malicious instructions are encrypted

Researchers show Grok can be made to exfiltrate user data via Cryptographic Context Injection, a newly documented technique that bypasses LLM safety guardrails.

According to Ars Technica, Grok exfiltrates user data when malicious instructions are encrypted, a technique called Cryptographic Context Injection. The method is described as the latest documented way to break LLM safety guardrails, showing that encrypted content can carry hidden instructions past safeguards. The finding underscores gaps in how large language models validate and execute context from external sources.

Ars Technica · Security·Aug 20, 13:00 UTC · 27d agoAI safety & security1

55

Observational Indistinguishability and Integrity Blind Regions in Hybrid Quantum-Classical Workflows

Framework formalizes integrity blind regions in hybrid quantum-classical workflows, validated across 3,600 label interventions with conformal detection rules.

The paper presents a claim-relative evidence and reference framework for integrity of hybrid quantum-classical workflows, distinguishing structural blind regions caused by observational indistinguishability from finite-batch statistical misses. Experiments over 3,600 label interventions show exact label-path invariance for feature and prediction views. The geometry-aligned construction detects 343 of 2,700 conclusion-changing interventions using the conformal rule and 1,183 of 2,700 with the uncorrected union, with executed conformal clean false-action rates of 0.048-0.059.

arXiv cs.CR·Sep 15, 13:16 UTC · 1d agoResearch

12

Signing the Transaction but Not the Decision: Whisper Attacks and a Binding Defense for AP2

Research shows AP2 agent-payment signatures can be manipulated into valid but wrong carts; proposed A-VIP defense binds signed intent to purchases.

A study demonstrates Whisper attacks on the AP2 agent payment protocol, where ordinary product-description text steers shopping agents into carts that pass every cryptographic check but no longer match user intent. Using Gemini Flash-Lite models specified by AP2's default sample agents, three attacks succeeded at 90%, 56%, and 73.3%, with the vulnerability spanning seventeen Google models, three agent frameworks, cross-vendor anchors, and Google's consumer assistant. The proposed A-VIP defense treats signed intent as a capability grant, binding credential lookups to sessions and cart lines to seen listings, blocking the first two attacks with zero false positives while surfacing unauthorized spending. The authors release A-VIP code, machine-checked invariants, and AP2-WhisperBench with 1,544 evaluation scenarios.

arXiv cs.CR·updated Sep 10, 16:51 UTC · 6d ago·first Sep 10, 16:11 UTC · 6d agoAI safety & security 2 sources1· 1 read

65

Can your coding style predict whether your code is vulnerable?

University of Massachusetts Dartmouth researchers present VulStyle, a stylometry-based vulnerability detector that also exposes benchmark reliability problems.

VulStyle combines stylometric features with syntax-tree structure and source tokens, pre-trained on about 4.9 million functions across seven programming languages and fine-tuned on five vulnerability detection datasets. It beat token-only detectors on some benchmarks but its F1 drops sharply on DiverseVul, which the authors link to noisy labels inflating reported performance across popular datasets. The authors argue style-aware detection should be harder to evade but did not test this empirically, and they note that uniform LLM-generated code may strip away the individual developer style the model depends on.

Help Net Security·Aug 24, 04:21 UTC · 23d agoResearch1

48

How Fragile Is Safety Alignment at Frontier Scale? A Single-Direction Attack on a 320B MoE

Researchers show directional ablation breaks refusal in GLM-5.3-Flash, a 320B-parameter MoE, cutting refusal by 41–89 points across seven benchmarks.

The study extends directional ablation, a white-box attack that removes an aligned LLM's refusal behavior, from dense models up to ~70B parameters to GLM-5.3-Flash, a 320B-parameter mixture-of-experts model with 288 routed experts, four-wide hyper-connection residual, and block-FP8 quantization. Editing attention, dense, and routed-expert writers jointly removes 0.776 of refusal, with 74% of the effect existing only under the joint intervention; the conventional module-name-based recipe reaches only 0.066 and fails silently on MoE architectures. The attack yields 41–89 percentage-point reductions in refusal across seven harmful benchmarks with no detected capability change, and a category-concentrated refusal residue survives all edits at ranks 1 to 12.

arXiv cs.CR·Sep 9, 06:48 UTC · 7d agoAI safety & security

45

I Am No One: Style-Aware Paraphrasing for Text Anonymization

Prompt-driven style-aware paraphrasing with LLMs cuts authorship attribution F1 by 60-70% while preserving content quality.

The paper proposes a style-aware, prompt-driven anonymization approach using pretrained LLMs to build compact stylistic profiles from minimal samples and rewrite text to suppress identifiable style markers while preserving meaning. It addresses stylometric re-identification risks in anonymized text, including ASR transcripts of meetings and call-center calls where leakage persists after acoustic anonymization. Across blog and review datasets, the approach reduces authorship attribution F1 by 60-70%, substantially outperforming both DP-based and non-DP baselines while maintaining readability.

arXiv cs.CR·Sep 11, 02:07 UTC · 5d agoResearch

40

[0day-rubbish] Accurate Online Private Cloud on-prem (current) Unauthenticated Hessian deserialization leading to JNDI remote class loading (9.8)

0day Rubbish disclosed an unauthenticated Hessian deserialization flaw in Accurate Online Private Cloud on-prem allowing JNDI remote class loading, rated 9.8.

The 0day Rubbish Research Team publicly disclosed an unauthenticated Hessian deserialization vulnerability in the current on-premises release of Accurate Online Private Cloud. The flaw lets unauthenticated attackers trigger JNDI remote class loading, a path that typically yields remote code execution. The issue carries a CVSS 9.8 rating. No CVE identifier or evidence of in-the-wild exploitation was included in the disclosure.

Full Disclosure·Sep 8, 17:31 UTC · 8d agoVulnerability

40

Reference-Based Bias Detection in LLMs via Relative Representations of Hidden States

Researchers propose auditing LLM bias via relative hidden-state representations, detecting bias increases with 3-50x less compute than output-level benchmarks.

The paper introduces a reference-based bias auditing method that compares hidden-state representations across model variants, such as before and after fine-tuning, by encoding sentences relative to a fixed anchor set. The resulting Representational Bias Shift (Delta-B) correlates with output-level bias change in 15 of 18 tested settings, reaching |r| = 0.84 under full fine-tuning across WildGuardMix, DecodingTrust, and ToxiGen benchmarks. Thresholding Delta-B detects checkpoints whose bias increased with ROC AUC between 0.65 and 0.99 and beats a SEAT-based baseline, while auditing a model in about three minutes with 3-50x less compute.

Hugging Face daily papers·Sep 8, 20:00 UTC · 8d agoAI research1

42

Beneath the Surface of Chains-of-Thought: A Mechanistic Interpretation of Reasoning Operations in LLMs

Study shows LLM reasoning operations like planning and deduction are geometrically separable in hidden states, with separability peaking in middle layers.

Researchers investigate whether functional reasoning operations — problem formulation, goal decomposition, deduction — have corresponding geometric structure in LLM hidden representations. They find operations are separable in held-out representations with separability peaking in middle layers, ruling out lexical and positional confounds; token-wise operation alignment becomes more distributed across layers, and identical surface tokens are represented differently depending on their surrounding chunk. Attention-masking interventions show chunk-onset operation-aligned representations depend on preceding reasoning context; code is released on GitHub (naver-ai/beneath-cot).

Hugging Face daily papers·Sep 3, 20:00 UTC · 13d agoAI research1

40

PIDS-Bench: Evaluating Prompt-Injection Detectors Under Over-Defense, Obfuscation, and Distribution Shift

PIDS-Bench shows prompt-injection detectors scoring F1 above 0.98 still misclassify about one-third of external benign security-adjacent prompts, revealing provenance-sensitive over-defense.

PIDS-Bench is a frozen multi-axis benchmark that jointly evaluates prompt-injection detectors on attack detection and benign false-positive behavior at fixed thresholds, spanning in-distribution inputs, hard-benign prompts, obfuscated attacks, and domain/structural distribution shifts. It evaluates seven detectors plus a rule-based lower-bound reference. A detector exceeding F1 = 0.98 on held-out data still misclassifies roughly one-third of an externally-sourced benign security-adjacent subset, and no internal detector reaches F1 >= 0.95 with hard-benign FPR <= 0.10 on the stress distribution. Hard-negative augmentation nearly eliminates over-defense on curated stress inputs but leaves it intact on externally-sourced prompts, a pattern termed provenance-sensitive over-defense.

arXiv cs.CR·Sep 14, 04:28 UTC · 2d agoAI safety & security

38