Updated daily124 verified stories
AI safety, security, alignment, evaluations, incidents, vulnerabilities, and safeguards tracked without hype.
Every story links to the strongest available evidence: original sources when available, otherwise clearly attributed reporting.
What happened, why it matters, and what to watch — no jargon tax.
When the signal is thin, we publish nothing rather than padding the feed.
A growing stream of verified perspectives for people who need to understand AI without chasing hype.
Anthropic says Claude Mythos 5 is now powering vulnerability scans in Claude Security for Enterprise customers, while partner integrations are planned. It is also creating a $35 million credit fund for open-source security and expanding its cyber-verification program.
NVIDIA argues that AI-agent security controls should be enforced by runtimes and infrastructure, rather than by modifiable model or harness logic. Its technical blog proposes a layered architecture and four workload profiles, but provides no independent validation of the approach.
The Nearby Glasses project uses Bluetooth data to flag some smart-glasses devices on Android and iOS. Its own documentation says the system can produce false positives, miss devices and should not be used to confront people.
Google Threat Intelligence Group says adversaries are using AI to accelerate vulnerability research, evade detection, automate malware actions and target the software supply chain around AI systems.
An arXiv preprint reports that language models may reveal sensitive data placed in context, even after refusing direct extraction requests. In controlled experiments, the authors say adaptive attacks recovered short secrets from ordinary outputs and extracted full Social Security numbers from a production-style agent.
A paper accepted by IEEE ICDM 2026 reports that adaptive inversion reconstructs substantially more of the original text from Gaussian-noise-protected embeddings than an existing generative baseline, challenging a widely used privacy assumption without establishing real-world exposure or independent replication.
An arXiv-listed Ph.D. dissertation studies how backdoor attacks can affect language models and vision-language models, including methods for analysis, detection and attack design.
An arXiv preprint introduces a weight-editing method intended to make safety refusals harder to extract and remove. The paper reports stronger post-abliteration refusal scores on two open models, with different tradeoffs in general-purpose performance.
A new approach to training multimodal large language models (MLLMs) eliminates the need for extensive task-specific supervision.
This work examines emoji-augmented prompts as a test case for gaps in safety evaluation of large language models (LLMs).
AWS says an expanded Policy Authoring feature can convert written rules into Dogwood policies that restrict agent actions over time, enforce prerequisites and limits, inspect free-form text with Bedrock Guardrails, and constrain tool inputs.
An arXiv paper introduces FraudBench, a benchmark for testing whether tool-using banking agents can detect fraud that unfolds across conversations. In a preliminary single-trial evaluation, four agents scored 49% to 65% on attack security.
One useful briefing each week
Get the week’s verified AI news, original data, useful tools, learning picks, and fresh AI jobs.
Hiring an AI professional or launching a useful AI product? Put it in front of people who came here to learn and act.
Post an AI job Submit an AI tool