Quick AI News - 10/06/2026 -Reflection AI's Beam, ChatGPT Ads, and AI Becoming Infrastructure
Learn AI in BitsShow Notes
This quick news update covers the AI developments from October 5 into October 6, 2026 worth knowing, from a new open-weight model to a critical security patch and a shift in how big tech companies use Anthropic's Claude internally.
Reflection AI released details of Beam, its first open-weight model: a mixture-of-experts design with 501 billion total parameters but only about 23 billion active per token. Reflection says Beam is competitive with models like GLM 5.2 on coding and reasoning while using roughly 3 to 4 times less inference compute, though that efficiency comparison is the company's own claim rather than an independently verified benchmark. Weights are expected to follow later in October under an Apache 2.0 license.
OpenAI announced a new advertising format for ChatGPT, which it says now reaches 1.2 billion people each week, with visual ads set to begin testing in the US later this month. The ads can appear alongside image generation, won't change model responses or user-generated images, and won't show for Plus, Pro, or Enterprise subscribers.
On security, GitLab patched a critical vulnerability in its self-hosted AI Gateway, CVE-2026-90970, with a CVSS score of 9.9, that could let an authenticated user with Duo Agent Platform access escape the prompt-template sandbox and execute commands on the Gateway. Separately, researchers using Anthropic's Mythos model identified a vulnerability in the open-source Rejetto HFS file server involving weak random-number generation, enabling session forgery and remote code execution, with attackers probing it shortly after disclosure.
On the business side, The Information reports Microsoft and Meta are both cutting internal use of Claude: Microsoft's expected internal Claude spending reportedly fell by more than a third, and Meta's internal Claude Code users reportedly dropped from about 60,000 to around 30,000, as both companies push employees toward their own AI tools. Separately, two former Groq engineers sued over Nvidia's roughly $20 billion 2025 deal with Groq, alleging the transaction favored insiders at other shareholders' expense.
Sources & References Reflection AI: Introducing Beam: Reflection's 501B open-weight model — https://reflection.ai/blog/introducing-beam OpenAI: Building advertising for the way people use AI — https://openai.com/index/new-chatgpt-ads-format-and-measurement/ GitLab: Critical Patch Release for Self-Hosted AI Gateway — https://docs.gitlab.com/releases/patches/other-patches/patch-release-gitlab-ai-gateway-19-4-1-released/ Horizon3: Anthropic's Mythos and Rejetto HFS — https://horizon3.ai/attack-research/disclosures/anthropic-mythos-rejetto-hfs-rce/ The Register: Anthropic's Mythos vulnerability story — https://www.theregister.com/security/2026/10/03/anthropics-super-bug-hunting-model-mythos-is-hardcore-good-at-math-as-latest-vuln-under-attack-shows/ The Information: Microsoft Slashes Internal Claude Spending by a Third — https://www.theinformation.com/articles/meta-microsoft-work-wean-staff-anthropics-claude Financial Times: Nvidia's $20bn Groq Deal Faces Lawsuit — https://www.ft.com/content/93ee425d-9ac7-4548-8cc9-fef2e0670787
Voice narration is AI-generated.
More from Learn AI in Bits
A wave of cybersecurity disclosures has made this week feel like a turning point, with people calling it "Black October." This episode treats it as a cluster of high-impact incidents, not proof that one AI-driven campaign caused them all, and separates where AI was directly involved from where the failures were conventional.
Recursive self-improvement is the idea that an AI can improve the very process it uses to improve itself, and this episode separates what's been demonstrated so far from the dramatic predictions often attached to the term. It draws a clear line between self-correction (revising one answer), automated optimization (keeping the best of several tested variants), and true recursive improvement, where the system changes the process that generates future improvements, so the next round starts with a better improver.
This week's AI News Sunday Wrap-Up covers a week where the industry moved on several fronts at once: new flagship and open models, enterprise agent governance, a serious safety warning, legal pressure, and the physical infrastructure limits behind the AI race.
Meta's Muse story is moving fast, and this episode takes a closer look at Muse Spark, the multimodal reasoning model family behind Meta's push into AI agents, including yesterday's release of an open-weight version for developers.
When a large language model generates text one token at a time, even a fast GPU hits a sequential bottleneck. This episode explains speculative decoding, an inference technique that prepares several tokens at once and checks them together, and when the extra computation pays off.