GAI Press.

OpenAI Sued Over Allegation That ChatGPT Drafted Lethal Poisoning Plan

Original headline: OpenAI Sued Over Claim ChatGPT Drafted Lethal Poisoning Plan

OpenAI was sued over allegations that its ChatGPT generated a detailed lethal poisoning plan. Plaintiff James Gates says he discovered a handwritten note listing toxic substances and dosages, suggesting his wife intended to poison him.

The case highlights the potential for generative AI to be misused for violent crimes, underscoring the need for regulation.

Research angleHow do harmful outputs from large language models affect users' perceived trust in AI? — Protection Motivation Theory

Anthropic Won't Patch Claude Dispatch Flaw Allowing Remote Commands Across Boundaries

Original headline: Anthropic won't fix a Claude Dispatch flaw that could let attackers execute remote commands across corporate boundaries

Anthropic announced it will not patch a vulnerability in its Claude Dispatch service. The flaw could let attackers execute remote commands across corporate boundaries, and can be exploited in six steps from privilege escalation to remote code execution (RCE).

The unpatched flaw could undermine inter‑enterprise security barriers, raising the risk of large‑scale breaches.

Research angleHow does the Claude Dispatch vulnerability affect organizations' willingness to adopt AI security controls? — Technology Acceptance Model (TAM)

Researchers Bypass ChatGPT Lockdown Mode to Exfiltrate Chats Without User Interaction

Security researchers disclosed a now‑patched vulnerability that allowed the Lockdown Mode in OpenAI’s ChatGPT to be bypassed, enabling silent extraction of conversation data without any user interaction. The flaw could have been abused for covert data theft, phishing and other attacks. OpenAI has released a patch to close the issue.

The find demonstrates that safeguards against prompt‑injection data leaks can be circumvented, raising concerns for AI‑driven privacy protection.

Research angleHow does the presence of Lockdown Mode bypass vulnerabilities affect organizations’ intention to adopt AI security controls? — Protection Motivation Theory

India CERT Warns AI-Driven 'Crafty Flacon' Group Targeting Fintech and Crypto Firms

Original headline: India Warns of New AI-Driven Cyber-Attacks Targeting Financial and Crypto Businesses

The Indian Computer Emergency Response Team (CERT-In) issued an alert that a newly identified threat actor, dubbed ‘Crafty Flacon’, is leveraging Chinese-made artificial-intelligence tools to launch attacks against financial-technology and cryptocurrency companies. The group employs AI-powered automation to discover vulnerabilities and deploy malicious code, aiming to infiltrate critical financial systems.

AI-powered attacks are rapidly raising the security risk for fintech and crypto sectors, prompting a need to reassess regulations and defensive measures.

Research angleHow does the use of AI tools by threat actors affect the success rate of attacks on financial institutions, and what role does Protection Motivation Theory play in shaping firms’ defensive investments?

Chainalysis warns all users' data exposed in cyber breach

Original headline: Chainalysis warns all users hit in cyber breach involving ‘agentic’ techniques

Blockchain analytics firm Chainalysis said on Wednesday night that a cyberattack exposed some information belonging to all its users, including those in law‑enforcement and national‑security sectors. The breach reveals that sensitive data from these entities may now be accessible to unauthorized parties.

The exposure of every user’s data undermines trust in blockchain analytics services and raises security concerns for law‑enforcement and national‑security agencies.

Research angleHow does a cyber breach influence users’ security compliance intentions? — Protection Motivation Theory

Security Industry Urges Governance of Enterprise AI Agents as Proxy Identities

Original headline: The Agent Smith Dilemma: Governing Agentic Identities

The security community notes that every autonomous AI agent deployed within an enterprise functions as a new proxy identity inside the network perimeter. Because these agents can act independently, organizations must establish identity management and control mechanisms for them.

Uncontrolled AI agents expand the attack surface, making robust governance essential for enterprise security.

Research angleHow does the adoption of autonomous AI agents affect employees' intention to comply with security policies? — Protection Motivation Theory

SecurityDataPipeline.com Launches Ungated Educational Resource for AI Era

Original headline: SecurityDataPipeline.com Launches as a New Educational Resource for the AI Era

SecurityDataPipeline.com has released an ungated educational resource aimed at the AI era. The guide explains how security data pipelines can lower SIEM ingestion costs, preserve detection coverage, and ready the data layer for AI-driven security operations.

The resource helps security teams lower SIEM ingestion costs while maintaining detection coverage as they adopt AI-driven operations.

Research angleHow does implementing security data pipelines affect SIEM cost reduction and detection efficiency? — Resource-Based View (RBV)

On‑path attack cracks Chainalysis tools, steals law‑enforcement tokens

Original headline: October crypto-tracker hack: Access cracked, law enforcement tokens packed

In October, an agentic on‑path attack compromised Chainalysis’s cryptocurrency‑tracking tools, breaking into the system and extracting the authentication tokens used by law‑enforcement agencies. The breach gave the attackers access to data that supports police investigations.

Compromise of tools that police rely on undermines crypto‑crime investigations and exposes sensitive enforcement credentials.

Research angleHow does a security breach of law‑enforcement crypto‑tracking tools affect investigative efficiency? — Protection Motivation Theory

Victoria's major parties clash over AI-generated videos in election ads

Original headline: Scare campaigns: Victoria’s major parties trade blows over AI-generated videos of leaders in election ads

Victoria’s major political parties exchanged accusations after AI-generated videos of their leaders appeared in election advertisements. Experts warned that such deepfakes could further erode public trust in politics. One example featured the Liberal leader warning that an ‘Americanised healthcare’ system could cost $150,000 for a heart‑attack treatment.

AI-generated deepfakes in campaign ads threaten political trust and democratic legitimacy.

Research angleHow do AI-generated election ads affect voter trust? — Technology Acceptance Model (TAM)