Tag: #ai

9th Circuit Vacates Injunction Against Perplexity, Says Users — Not Perplexity — Accessed Amazon

A Ninth Circuit panel vacated an injunction against Perplexity’s Comet assistant, ruling the record shows users — not Perplexity — accessed Amazon.

$AMZN, #amazon, #perplexity, #cfaa, #ai, #ninthcircuit

Researchers Find Position-Independent KV Cache Reuse Can Let One User Hijack Another’s LLM Output

Researchers show position-independent KV-cache reuse in LLM inference can let attackers' context influence other users' outputs, posing risks for shared serving.

#ai, #security, #infrastructure, #ml, #llm

Amazon says custom chips surpass $25 billion annual run rate as Trainium and Graviton power AWS AI

Amazon reports its custom-silicon business topped a $25B annualized run rate, citing Trainium, Graviton and multi-year AI customer commitments.

$AMZN, #amazon, #aws, #ai, #chips

EU to Begin Enforcing AI-Transparency Rules on Aug. 2, 2026

From Aug. 2, companies serving the EU must disclose AI interactions and mark synthetic audio, image, video and text under Article 50 — violations face heavy fines.

#ai, #eu, #regulation, #transparency

AI agents can run experiments and draft papers but fail to produce publishable research, study finds

A new preprint shows AI agents can handle engineering tasks—code, experiments and paper drafts—but failed to make publishable advances in two NeurIPS case studies.

#ai, #research, #machinelearning, #neurips

Samsung Posts Record Q2 Earnings as AI Memory Chips Drive Profits

Samsung posted record Q2 revenue and operating profit, driven almost entirely by booming AI memory-chip sales (HBM4) while phones lagged.

#samsung, #semiconductors, #ai, #earnings

Preprint: Larger LLMs May 'Snowball' Hallucinations Faster in Long Factual Answers

A July arXiv preprint finds larger language models start stronger but may compound hallucinations faster within long factual responses.

#ai, #language-models, #hallucinations

Meta signs EU transparency code for AI‑generated content ahead of Aug. 2 rules

Meta signed the EU's voluntary Code of Practice on transparency for AI‑generated content, seeking a clearer path to comply before Aug. 2.

$META, #meta, #ai, #eu, #transparency

Oxford preprint finds GraphCast and NeuralGCM fail basic coordinate-change tests

An Oxford preprint on arXiv reports GraphCast and NeuralGCM fail latitude/longitude flips and rotations, raising questions about physical generalization.

#ai, #weather, #climate, #ml

White House Announces Expansion of Genesis Mission, Citing More Than $5 Billion in Federal Commitments

On July 22 the White House expanded the DOE-led Genesis Mission, citing more than $5 billion in federal commitments and 278 selected AI-for-science projects.

#ai, #genesismission, #departmentofenergy, #federalpolicy

Google DeepMind Expands Gemini Flash Lineup with 3.6 Flash and Security-Focused 3.5 Variant

Google DeepMind added Gemini 3.6 Flash plus two 3.5 variants, including a restricted 'Flash Cyber' model for defensive vulnerability discovery and repair.

$GOOGL, $GOOG, #google, #ai, #gemini, #cybersecurity

Microsoft Research unveils GigaPath-Flash and GigaTIME-Flash — smaller open pathology AI models that cut compute and memory

Microsoft Research and collaborators release two smaller, open pathology AI models that reduce compute and memory while keeping most performance.

#ai, #pathology, #medicalimaging, #openmodels

Chained Harmless Tools Let LLM Agents Perform Harmful Real-World Actions, Study Finds

An arXiv preprint shows attackers can chain innocuous tool calls to make LLM agents carry out harmful operations, with about 91% success.

#ai, #security, #machinelearning, #agents

Academic Paper Details “Action Rebinding” Flaw That Can Trick Android GUI Agents

Academic paper describes “Action Rebinding,” a TOCTOU-style exploit that can trick Android GUI agents into acting in other apps during seconds-long reasoning delays.

#android, #security, #ai, #vulnerabilities, #research

White House launches GOLD EAGLE clearinghouse to speed patching of open-source and critical-infrastructure vulnerabilities

The White House unveiled GOLD EAGLE, a Treasury-led clearinghouse aimed at accelerating reporting, validation and patching of software flaws across open source and critical infrastructure.

#cybersecurity, #ai, #open-source, #whitehouse