Just in:
Yeebo Announces Interim Results for FY26/27 Revenue Increases by 45.5% to HK$806.7 Million Profit Attributable to Owners Surges 87.6% to HK$2,284.7 Million // Anthropic shares finish above offer price after Nasdaq debut // International Business Leaders Say Xi’s Speech Boosts Confidence in Asia-Pacific Cooperation // Dar Global unveils $700m Trump development in Qatar // Travel Insurance Was Built for the Few Who Claim. Ancileo’s Felt Coverage™ is for Everyone Else // European Commission takes systems offline after AI cyberattack // Russia permits retail cryptocurrency purchases under regulated framework // LittleLives, Little Lab and SEGi College Sign MOU to Explore Early Childhood Education Collaboration Across ASEAN // Sabarimala booking discrepancies prompt sabotage angle probe // Workday targets Honeydew purchase to unify AI metrics // Diabetes Malaysia And Alpro Pharmacy Set National Record With 11,007 Most Blood Glucose Screenings Conducted Across Malaysia In Just 48 Hours // Diabetes Malaysia And Alpro Pharmacy Set National Record With 11,007 Most Blood Glucose Screenings Conducted Across Malaysia In Just 48 Hours // Diabetes Malaysia And Alpro Pharmacy Set National Record With 11,007 Most Blood Glucose Screenings Conducted Across Malaysia In Just 48 Hours // Padmini follows Padmaja for ‘ghar vapsi’ // Balagopal challenges CM over tax revenue // Phancy Announces Business Performance for the Nine Months Ended September 30, 2026 // Sharjah clears Kalba free zone plan for 400MW complex // PayerMax and ShopeePay Strengthen Strategic Partnership to Advance Digital Payment Collaboration in Southeast Asia // Sui activates Hashi Bitcoin lending across partner network // Oil futures rebound after Riyadh strikes and tanker alerts //

LLM guardrails falter under dialogue attacks

Cisco researchers have warned that leading open-weight large language models can be manipulated through sustained conversations that gradually push them past safety controls, exposing a weakness in systems now being adopted across business, public services and consumer applications.

The assessment tested eight widely used open-weight models from Alibaba, DeepSeek, Google, Meta, Microsoft, Mistral, OpenAI and Zhipu AI. The models were examined through automated adversarial testing designed to measure whether they could resist prompt-injection and jailbreak attempts across both single-turn and multi-turn exchanges.

The findings point to a marked gap between how models behave when challenged with one direct prompt and how they respond when harmful intent is introduced over several conversational steps. Multi-turn attacks achieved success rates ranging from 25.86 per cent to 92.78 per cent, with some models proving two to 10 times more vulnerable in extended dialogue than in single-prompt tests.

The risk is significant because many enterprise AI systems are built around chat interfaces, agents and assistants that depend on long exchanges with users. A request that would be blocked if made directly may be broken into smaller, apparently harmless steps, allowing the user to build context, establish a role-play scenario or gradually steer the system towards prohibited output.

Cisco’s researchers described the pattern as a systemic weakness in the ability of current open-weight models to maintain safety instructions across longer conversations. The tests were conducted as black-box engagements, meaning the internal architecture and any additional safety layers were not disclosed before assessment.

The models tested included Qwen3-32B, DeepSeek v3.1, Gemma 3-1B-IT, Llama 3.3-70B-Instruct, Phi-4, Mistral Large-2, GPT-OSS-20b and GLM 4.5-Air. The research did not argue against open-weight AI development, but said organisations need to understand the security posture of models before using them in production or fine-tuning them for sensitive tasks.

Open-weight models have become central to the AI ecosystem because they allow developers to inspect, customise and deploy systems without relying entirely on closed commercial platforms. Their growth has accelerated across research, software development, cyber security operations, customer service and internal knowledge tools. That flexibility also creates exposure when models are deployed without layered protections.

Capability-focused models showed larger gaps between single-turn and multi-turn performance, while models with stronger safety alignment appeared to perform more consistently across attack types. The distinction matters for enterprises choosing systems not only for speed, cost or benchmark performance, but also for resilience against manipulation.

Security specialists have warned that model capability benchmarks often overshadow safety testing. A model that performs well in coding, reasoning or language tasks may still be weak against adversarial dialogue. This creates a procurement risk for organisations that select models on productivity metrics while underestimating misuse scenarios.

The concerns extend beyond harmful text generation. Multi-turn manipulation could affect systems connected to databases, code repositories, workflow tools, customer records or decision-support platforms. A compromised AI assistant could expose confidential information, generate misleading material, alter business logic or assist in unauthorised activity if linked to operational systems.

The threat becomes sharper as AI agents gain the ability to take actions rather than merely produce text. When models are connected to tools, calendars, cloud environments, ticketing systems or financial workflows, a successful jailbreak may have consequences beyond the chat window. Guardrails therefore need to monitor not only individual prompts but the full conversational trajectory.

Researchers in the wider AI safety field have also found that multi-turn attacks are harder to detect because each message can look benign when viewed alone. The malicious intent becomes clear only when the dialogue is assessed as a sequence. That creates a challenge for filters that operate at the level of isolated inputs and outputs.



Notice an issue?

Arabian Post strives to deliver the most accurate and reliable information to its readers. If you believe you have identified an error or inconsistency in this article, please don't hesitate to contact our editorial team at editor[at]thearabianpost[dot]com. We are committed to promptly addressing any concerns and ensuring the highest level of journalistic integrity.


Loading next story…
Just in:
ED searches five districts in probe against hawala // Yeebo Announces Interim Results for FY26/27 Revenue Increases by 45.5% to HK$806.7 Million Profit Attributable to Owners Surges 87.6% to HK$2,284.7 Million // Jaiedco and Golden Bridge Announce Strategic Collaboration for a New Dubai Islands Development // Dubai Introduces SHARH, an AI-Powered Platform Helping Entrepreneurs Buy, Sell and Manage Businesses in the UAE // LittleLives, Little Lab and SEGi College Sign MOU to Explore Early Childhood Education Collaboration Across ASEAN // Sui activates Hashi Bitcoin lending across partner network // Travel Insurance Was Built for the Few Who Claim. Ancileo’s Felt Coverage™ is for Everyone Else // Diabetes Malaysia And Alpro Pharmacy Set National Record With 11,007 Most Blood Glucose Screenings Conducted Across Malaysia In Just 48 Hours // PayerMax and ShopeePay Strengthen Strategic Partnership to Advance Digital Payment Collaboration in Southeast Asia // Dar Global unveils $700m Trump development in Qatar // Thailand’s Most Iconic New Year’s Celebration Returns: ICONSIAM Unveils “Amazing Thailand Countdown 2027” // European Commission takes systems offline after AI cyberattack // Balagopal challenges CM over tax revenue // Sabarimala booking discrepancies prompt sabotage angle probe // Padmini follows Padmaja for ‘ghar vapsi’ // Dalilk Academy reaches 27 million, targets 100 million by 2030 // Toward a Shared Future for Humanity: Rethinking Civilizational Exchange // Crypto market loses $150 billion amid broad sell-off // Phancy Announces Business Performance for the Nine Months Ended September 30, 2026 // Anthropic shares finish above offer price after Nasdaq debut //