standard
OAuth 2.0 is an open standard for delegated authorization, letting apps obtain scoped access tokens to act on a user's behalf without sharing passwords.
No evidence-backed relationships are recorded.
A dev.to post says Sign in with ChatGPT charges an app's AI requests to the user's own token budget, but only for users who link an account. Users who skip the link or run out of quota fall back to a cheaper model that the developer still pays for.
Reality
MCP Python SDK maintainers fixed a flaw rated up to 7.5 that lets malicious servers steal OAuth client secrets, in versions 1.30.0 and 2.2.0. Two of the affected providers stay exposed after upgrading until the calling code passes issuer=.
Reality
Enterprise AI tooling now trips about one SOC alert in 230, the fastest-growing class in the stream, and 94.1% of what it fires is legitimate work hitting detection rules written before agents existed. The bill is tuning and triage capacity.
Reality
F5 disclosed CVE-2026-94127 on September 22 with hotfixes and evidence of exploitation. It is a data plane heap overflow, so a locked-down management interface still leaves the system exploitable, and federal agencies had until September 25.
Perspective Coverage
6 publishersReality
CSO Online's teardown of a Webworm implant found an upgrade command that replaces all five credential strings in a single task. The revocation step at the top of most identity runbooks costs the operator one poll cycle.
Reality
Agent SSO went generally available on August 24 at no extra cost inside core Okta SSO, registering Cross App Access agents in Universal Directory and issuing short-lived tokens. Discovering unregistered agents takes a separate subscription.
Reality
A developer suspended a signed-in test user while that user's Claude agent was still running, and the app kept reading the token as valid, so enforcement has to move into a status check the application itself owns.
Reality
A Help Net Security column reads the two intrusions as one sequence run twice, with the token first and the inbox second. It says the same four steps describe what an authorized AI agent does in Google Workspace every day.
Reality
Microsoft's guidance on illicit consent grants says password resets and MFA do not touch the access, and it sends admins to the Purview audit log, where the entry can take up to 24 hours to appear.
Reality
The FBI says a subscription kit sold on Telegram harvests Microsoft 365 OAuth tokens through device code lures. The mitigation it recommends is a tenant-wide block, and somebody has to decide the exceptions.
Reality
EarthLink Network's in-house identity platform now settles administrator status in one function, on both token issue and refresh, using the group-name check its products were already applying, and the UI only displays the answer.
Reality
The spec's own boundary list says a verified catalog entry establishes which domain claims a capability and what trust material it published. It says nothing about whether the calling process is the agent it names.
Reality
Mitiga Labs says an npm post-install hook can repoint Claude Code's MCP endpoints at attacker infrastructure while the SaaS provider keeps logging a real user session arriving from Anthropic's egress range.
Reality
The SANS 2026 survey puts 74 percent of businesses on standing credentials for autonomous AI, with no single control above 40 percent. The two MCP incidents most often cited as proof have different root causes.
Reality
CVE-2026-14894 gives an unauthenticated attacker code execution on a WordPress site, and the fix is Super Forms 6.3.314. Microsoft separately reports invisible Unicode tag characters at up to 2.37 million messages a day.
Reality
A new Iceberg REST Catalog addition lets the catalog evaluate policy for one principal and return nine column actions plus row filters for a trusted engine to apply, while the spec leaves establishing that trust to the deployment.
Reality
A dev.to post counts seven safety controls MCP leaves to whoever integrates it. The catalog filter it recommends turns every advertised tool into a per-session authorization decision somebody has to own.
Reality
Plain API keys leak more than any other credential and still have no standard kill switch, so a draft called ORKS copies OAuth's discoverable revocation endpoint and lets the person who found the key pull it.
Reality
Grant de Swardt watched his Claude Max 20x allowance climb on a day he did no work. He asked Anthropic for an itemized usage log, and what he got was a two-week suspension and a partial refund of £44.49.
Reality
Reco says four in five AI tools in its telemetry run outside IT oversight, averaging 414 per 1,000 staff at smaller companies, which turns the cleanup into a question of who now owns each agent's credentials.
Reality
Build · August 29, 2026 · 1 publisher
Build · August 29, 2026 · 1 publisher
Build · August 28, 2026 · 1 publisher
Build · August 27, 2026 · 1 publisher
Build · August 25, 2026 · 1 publisher
Product · August 24, 2026 · 1 publisher
Build · August 24, 2026 · 1 publisher
Invest · August 24, 2026 · 1 publisher
Build · August 24, 2026 · 1 publisher
Build · August 23, 2026 · 1 publisher
Build · August 23, 2026 · 1 publisher
Build · August 22, 2026 · 1 publisher
Build · August 20, 2026 · 1 publisher
Build · August 20, 2026 · 2 publishers
Build · August 19, 2026 · 1 publisher
Security · August 14, 2026 · 1 publisher