Reflection releases Beam: a 501B open-weight coding model under Apache 2.0
Beam makes a large agent-focused model commercially reusable without a bespoke licence, but the open weights are not a small local download.
Search
Search results for “prompt injection”
Find published research by company, product, platform or technology.
Showing 41–60 of 115 dossiers
Gemini CLI 0.60 hardens the boundaries around extensions, sandboxes and MCP OAuth
This is not one headline vulnerability fix. Gemini CLI 0.60 is a coordinated hardening pass across the plumbing that lets extensions, sandboxes, filesystem paths and MCP authentication influence an agent’s execution environment.
Muse Spark 1.3 improves agent efficiency without raising Meta’s API price
Muse Spark 1.3 is more than a routine model refresh: Meta is pairing stronger agent behavior with lower vendor-reported tool/token use at the same published unit price. Independent testing supports a capability gain, but max reasoning can consume substantially more reasoning tokens.
Artificial IntelligenceWeb DevelopmentCloud & InfrastructureAI AgentsCloud PlatformsDeveloper ToolsDatabases & StorageActive dossier
Supabase makes MCP access centrally governed through enterprise SSO
Supabase has implemented MCP Enterprise-Managed Authorization using identity-provider assertions, short-lived tokens and existing Supabase role boundaries. It gives organizations a central on/off switch for approved AI clients while keeping access scoped to the individual employee rather than sharing a powerful organization token.
Artificial IntelligenceWeb DevelopmentCloud & InfrastructureDeployment & DevOpsInference & APIsDeveloper ToolsActive dossier
LiteLLM 1.107.0 was backdoored on PyPI and Docker: startup hooks stole credentials even without an import
The unusual risk is that Python startup hooks execute without an application importing LiteLLM, while the public source tag lacked the malicious files. Check the exact installed wheel or image digest, not just the repository.
pgBackRest 2.60 makes corrupt-page checks block restores and backup expiry by default
A safer default can surprise an incident-response runbook: pgBackRest 2.60 aborts restores with invalid page checksums and refuses certain backup expirations. Here's what operators should test.
Shopify’s non-Plus checkout deadline has passed — legacy Thank you and Order status scripts are sunset
The non-Plus checkout migration is now an active compatibility boundary rather than an approaching deadline. Orders can continue while old post-purchase scripts, pixels or widgets stop working, making end-to-end conversion and app-behavior checks important after the cutover.
GitHub Copilot adds model-judged permissions to its GA agent sandbox — with unattended-run traps
Copilot now offers model-reviewed action approvals between manual and Allow All. Unlike OS sandboxing, this probabilistic gate can approve unsafe operations; unattended CLI flags may leave unexpectedly broad permissions.
OpenAI’s Decisions API turns GPT-6 Luna into a bounded decision layer for software
Jev, CLM and GLiNER2.5-Decide made bounded software decisions look like a distinct model category. OpenAI is now validating the same architectural split with a Luna-powered API designed to answer finite questions rather than generate open-ended prose.
Google is turning Managed Agents into scheduled workers with budgets and hooks
The May Antigravity agent ID is retired. Managed Agents now require the September preview ID and default to Gemini 3.8 Flash, alongside hooks, token budgets and scheduled sandboxes.
Cognition’s Fusion uses a frontier lead and cheaper sidekick to cut coding-agent task cost
Fusion is interesting less as another routing feature than as a different agent-cost architecture: two persistent model contexts divide planning, review and execution instead of making one expensive model handle every token. The practical question for builders is shifting from token price to cost per completed task.
Gemini Omni Flash goes GA with video extension, interpolation and 4K output controls
Gemini Omni Flash has crossed from preview into a production API with a broader editing surface. Builders can extend existing clips and interpolate between images, but preview integrations now have a September migration deadline.
Artificial IntelligenceWeb DevelopmentJavaScriptAI CodingBrowsers & Web StandardsDeveloper ToolsActive dossier
Google's AI bug hunter finds different V8 flaws from fuzzing — Project Zero says keep both
A six-month real-world comparison shows AI code-review agents and specialised fuzzers find substantially different V8 vulnerabilities. Replacing fuzzing with agents would leave gaps, and cost per bug remains unknown.
GitLab patches a CVSS 10.0 CI/CD regex RCE — and five more high-severity flaws
CVE-2026-97229 carries GitLab's maximum 10.0 CVSS score. The 19.3 patch line requires 19.3.5, not the initially issued 19.3.4, because of an Ona integration regression.
Go's gopls adds interactive refactoring and proposes a new cross-editor LSP capability
gopls 0.24.0 and VS Code Go 0.56.0 ship interactive refactorings for struct tags and interface implementations; the broader Language Server Protocol extension is still a proposal, with cross-file moves in preview.
Google Preferred Sources now influences publisher visibility in AI Overviews and AI Mode
Google is moving source links higher in AI answers, testing prominent official-site panels and expanding news carousels, while Preferred Sources use has grown more than sevenfold in 2026.
Go 1.27 quietly makes HTTP/1 clients faster by reusing connections after early Close
Go 1.27 automatically drains up to 256 KiB or 50 ms after an HTTP/1 response body is closed, enabling reuse without custom drain code.
Artificial IntelligenceCloud & InfrastructureAI ModelsCompute & AI InfrastructureInference & APIsActive dossier
OpenAI's GPT-6.1 Sol Ultrafast tier now has public 6× Standard pricing
OpenAI's Ultrafast tier has moved beyond its GPT-5.6 preview: GPT-6.1 Sol access is broad and pricing is documented at 6× Standard.
ChartMogul finds the first 30 days dominate SaaS upgrades — across 10.5 million purchases
New SaaS cohort data challenges the habit of waiting six months to pitch an upgrade. The strongest seat and plan expansion window is the first month, while year-one renewal creates a second chance; AI-native customers are more likely to reactivate after churn.
Artificial IntelligenceWeb DevelopmentIndie BusinessAI CodingTiny TeamsDeveloper ToolsActive dossier
Rashomon gives Claude Code sessions a second, independent execution record
Rashomon's experimental local recorder can expose discrepancies between a coding agent's closing claims and its tool execution. It is an observability aid, not a sandbox or tamper-proof security product.