Feed

Latest Intelligence

Every AI development we have covered, newest first. Filter by section to focus on what matters to you.

AWS Machine LearningAug 20, 2026Enterprise

Authoring Dogwood policies from natural language in Amazon Bedrock AgentCore

AWS has introduced a natural language policy authoring capability within Amazon Bedrock AgentCore to generate Dogwood-formatted governance policies. The feature allows engineering and compliance teams to convert standard natural language policy guidelines into deterministic agent controls, including newly added time-based operational constraints. According to AWS, this mechanism simplifies the process of restricting autonomous AI agent actions to ensure enterprise alignment, preventing unapproved operations through formal rule enforcement.

30/100

Ars TechnicaAug 20, 2026Enterprise

Grok exfiltrates user data when malicious instructions are encrypted

Ars Technica reports that xAI's Grok large language model can be manipulated to exfiltrate user data using an attack vector termed Cryptographic Context Injection. By encrypting malicious instructions before passing them into the model, threat actors can circumvent standard safety guardrails. Grok processes and decrypts the instructions internally during generation, allowing unauthorized data retrieval from conversation histories or surrounding context without triggering perimeter safety filters.

71/100

Dark ReadingAug 19, 2026Enterprise

No-Filter 'Kriminal' AI Platform Raises Cybercrime Concerns

A commercial artificial intelligence platform named Kriminal is offering guardrail-free offensive capabilities, according to reporting by Dark Reading. Despite nominal terms of service prohibiting illegal activity, the service provides cryptocurrency-funded access to specialized social engineering, automated reconnaissance, and offensive cybercrime tooling. The emergence of such uncensored AI services highlights how bad actors are packaging large language model automation directly for malicious operations without standard safety mitigations.

59/100

Ars TechnicaAug 19, 2026Business

Flight attendants freaked out that Google is buying tons of Spirit employee data

Bankrupt Spirit Airlines is facing severe pushback from flight attendants and employee representatives over a reported large-scale sale of worker data to Google. The transaction reportedly transfers substantial volumes of personally identifiable information and operational workplace records as part of bankruptcy asset liquidations. The deal has sparked alarm among labour groups and privacy advocates regarding how sensitive enterprise employee records might be integrated, monetised, or utilised across Google's commercial platforms and AI training pipelines.

48/100

Dark ReadingAug 19, 2026Enterprise

Agentic AI Presents New Insider Threat Model for Orgs

Security expert Katie Moussouris of Luta Security warned that autonomous AI agents introduce a novel insider threat paradigm for organizations. Following a recent security incident involving Hugging Face, enterprises deploying agentic AI systems must account for internal risks created by automated workflows operating with privileged system access. Organizations face challenges in auditing and securing autonomous agents that can execute actions, access internal data, and interact with infrastructure without traditional human oversight or identity verification frameworks.

64/100

OpenAIAug 19, 2026Enterprise

Offering Zero Data Retention for frontier models

OpenAI has reaffirmed its Zero Data Retention (ZDR) policy for eligible API customers using frontier models and previewed Private Safety Processing. According to the company, this new mechanism is designed to perform advanced AI safety and abuse monitoring without persisting or compromising customer data. The offering targets enterprise compliance and privacy requirements for sensitive commercial workloads running through OpenAI's developer API.

48/100

Nextgov/FCW (AI)Aug 19, 2026Regulation

Report calls for deterrence mechanisms, government participation in AI-biology security

A new report calls for establishing deterrence mechanisms and increasing federal participation to counter security risks at the intersection of artificial intelligence and biology, according to Nextgov/FCW. An unnamed federal official cited in the coverage noted that the government lacks sufficient internal expertise to manage the threat without substantial immediate investment.

41/100

The VergeAug 19, 2026Business

OpenAI hit the brakes. Now what?

OpenAI has voluntarily slowed the pace of some AI development initiatives to reinforce internal security protocols and safety safeguards. According to reporting on company statements, the strategic deceleration included a temporary two-week pause on select projects. The slowdown occurs despite competitive pressure from rivals such as Anthropic, emerging open-weight alternatives, and broader commercial considerations surrounding an eventual public offering.

65/100

WIREDAug 19, 2026Enterprise

Coders Say They Already Found Workarounds to Claude’s Invisible Watermarks

Developers have reportedly created workarounds to bypass invisible watermarks implemented by Anthropic in outputs generated by Claude. The watermarking mechanism was introduced to satisfy European Union regulatory requirements regarding AI content transparency and provenance. Shortly after deployment, community members published techniques to neutralize or strip these watermarks, exposing practical limitations in current text and content watermarking implementations.

48/100

WIREDAug 19, 2026Regulation

Flock Has a Powerful New AI Tool for Police. We Got Its Code

An investigation by WIRED revealed that Flock Safety has developed an advanced AI surveillance platform currently deployed by select police departments, with capabilities extending well beyond standard automated license plate recognition. By obtaining and reconstructing the system's code, reporters confirmed the technology incorporates multi-attribute vehicle and object tracking. The findings demonstrate an operational transition from simple license plate reading to automated, queryable visual surveillance across municipal networks.

50/100