Explainer-How Ukraine’s strikes on Russian energy sites deepened the rift with Trump
investing.com/news/commoditi…
#UkraineWar #RussiaEnergy #OilMarkets #Trump #Geopolitics $USO $XLE $CL $OXY $CVX
investing.com/news/commoditi…
#UkraineWar #RussiaEnergy #OilMarkets #Trump #Geopolitics $USO $XLE $CL $OXY $CVX
0
0
0
26
Saudi Arabia intercepts drones and missiles as fighting with Yemen’s Houthis continues
investing.com/news/commoditi…
#SaudiArabia #YemenConflict #Houthis #MiddleEastTensions #OilMarkets
$USO $XLE $LMT $RTX $NOC
investing.com/news/commoditi…
#SaudiArabia #YemenConflict #Houthis #MiddleEastTensions #OilMarkets
$USO $XLE $LMT $RTX $NOC
0
0
0
41
El Salvador president says bodies found in Haiti do not belong to ambushed soldiers
investing.com/news/commoditi…
#ElSalvador #Haiti #SecurityCrisis #MilitaryNews #CaribbeanNews
$BTC $ETH $GOLD $OIL
investing.com/news/commoditi…
#ElSalvador #Haiti #SecurityCrisis #MilitaryNews #CaribbeanNews
$BTC $ETH $GOLD $OIL
0
0
0
239
Can silver recover from its latest sell-off? UBS weighs in
investing.com/news/commoditi…
#Silver #PreciousMetals #Commodities #MarketOutlook #UBS $XAGUSD $SLV
investing.com/news/commoditi…
#Silver #PreciousMetals #Commodities #MarketOutlook #UBS $XAGUSD $SLV
0
0
0
65
Iran has ramped up efforts to target US personnel, intelligence memo warns
investing.com/news/commoditi…
#Iran #USMilitary #MiddleEastTensions #Geopolitics #NationalSecurity
$USO $XLE $LMT $RTX $NOC
investing.com/news/commoditi…
#Iran #USMilitary #MiddleEastTensions #Geopolitics #NationalSecurity
$USO $XLE $LMT $RTX $NOC
0
0
0
29
Iran’s Pezeshkian blames US for interim deal failure, urges new try
investing.com/news/commoditi…
#IranNuclearDeal #USIranRelations #OilMarkets $USO $XLE
investing.com/news/commoditi…
#IranNuclearDeal #USIranRelations #OilMarkets $USO $XLE
0
0
0
12
Agentic AI is making the cyber control plane, not model behaviour alone, the limiting factor for assurance. The decisive question is increasingly whether identities, entitlements, asset records and vulnerability evidence can reliably govern machine action at runtime. Once an agent can act, weak control-plane data can turn a plausible recommendation into an authorised, repeatable mistake.
NIST’s warning that early agentic deployments are prioritising immediate value over security, and its emphasis on a strong identity foundation, points directly to this shift. Identity is no longer merely a user-access concern; it becomes the mechanism for defining what an autonomous workload is, what it may access, how far it may act, and when its authority can be withdrawn.
The NCSC’s observation that defenders cannot use AI in the same way attackers can reinforces the asymmetry. Attackers need only exploit a weakness, while defensive agents require trustworthy asset visibility, permissions, telemetry and escalation paths before they can safely make changes. Its call for input on external attack-surface management is therefore significant: an agent cannot safely remediate what the organisation cannot authoritatively identify.
NIST’s consultation on AI-enabled vulnerability management adds a second layer. AI may accelerate vulnerability discovery, prioritisation and compliance automation, but it also makes the provenance and freshness of vulnerability intelligence a security boundary. Recent NCSC alerts on actively exploited Citrix flaws and critical PHP issues show why: the value of automation rises precisely as the cost of stale, incomplete or wrongly scoped evidence rises.
Conventional assurance still overweights model testing and red-teaming. The emerging requirement is decision-to-action lineage: proof of the agent identity, authority, evidence sources, approvals, actions and revocation conditions behind every consequential change.
This trend will be confirmed if organisations standardise workload identities, fine-grained permissions, evidence provenance and action-level audit trails. It will be challenged only if broad, static privileges prove safe for autonomous operations, an increasingly implausible proposition.
#AgenticAI #AIsecurity #IdentitySecurity #CyberAssurance
NIST’s warning that early agentic deployments are prioritising immediate value over security, and its emphasis on a strong identity foundation, points directly to this shift. Identity is no longer merely a user-access concern; it becomes the mechanism for defining what an autonomous workload is, what it may access, how far it may act, and when its authority can be withdrawn.
The NCSC’s observation that defenders cannot use AI in the same way attackers can reinforces the asymmetry. Attackers need only exploit a weakness, while defensive agents require trustworthy asset visibility, permissions, telemetry and escalation paths before they can safely make changes. Its call for input on external attack-surface management is therefore significant: an agent cannot safely remediate what the organisation cannot authoritatively identify.
NIST’s consultation on AI-enabled vulnerability management adds a second layer. AI may accelerate vulnerability discovery, prioritisation and compliance automation, but it also makes the provenance and freshness of vulnerability intelligence a security boundary. Recent NCSC alerts on actively exploited Citrix flaws and critical PHP issues show why: the value of automation rises precisely as the cost of stale, incomplete or wrongly scoped evidence rises.
Conventional assurance still overweights model testing and red-teaming. The emerging requirement is decision-to-action lineage: proof of the agent identity, authority, evidence sources, approvals, actions and revocation conditions behind every consequential change.
This trend will be confirmed if organisations standardise workload identities, fine-grained permissions, evidence provenance and action-level audit trails. It will be challenged only if broad, static privileges prove safe for autonomous operations, an increasingly implausible proposition.
#AgenticAI #AIsecurity #IdentitySecurity #CyberAssurance
1
0
0
57
Ukrainian strike on enterprise in Russia’s Oryol kills three, governor says
investing.com/news/commoditi…
#UkraineWar #RussiaUkraineConflict #Oryol #DroneStrike #Geopolitics
$LMT $RTX $XAR $ITA
investing.com/news/commoditi…
#UkraineWar #RussiaUkraineConflict #Oryol #DroneStrike #Geopolitics
$LMT $RTX $XAR $ITA
0
0
0
23
Marco Rubio calls on those seeking change in Cuba to partner with CIA
investing.com/news/commoditi…
#Cuba #MarcoRubio #CIA #USForeignPolicy #CubaProtests
$USD $OIL $SPY
investing.com/news/commoditi…
#Cuba #MarcoRubio #CIA #USForeignPolicy #CubaProtests
$USD $OIL $SPY
0
0
0
29
The emerging constraint on agentic cyber defence is not model intelligence but delegated authority. As AI is positioned to discover, triage and act on security work, assurance is becoming a question of which identity may exercise which permission, against what evidence, and with what containment when it is wrong.
NIST’s warning that early agentic deployments are prioritising features and immediate value over security identifies the core architectural gap: agents require an identity foundation before they can be trusted with consequential enterprise actions. At the same time, NIST is seeking input on AI-enabled vulnerability management for the NVD, including automation opportunities in a field where data drives compliance and remediation. The implication is significant. AI may accelerate the conversion of vulnerability intelligence into action, but the risk is no longer merely an inaccurate finding. It is an agent acting on inaccurate, incomplete or adversarially influenced evidence.
The NCSC’s observation that defenders cannot use AI in the same way as attackers sharpens this asymmetry. Attackers need only exploit an opening; defenders must establish authority, preserve availability and justify intervention. Recent NCSC calls for immediate action on PHP and actively exploited Citrix NetScaler vulnerabilities show why pressure to automate will intensify. Yet urgency cannot substitute for accountable execution.
Conventional AI-security thinking still treats autonomy as a graduated capability problem: begin with a copilot, then permit more actions. The more important distinction is between reversible assistance and authority-bearing operations. An agent that drafts a remediation plan is fundamentally different from one that changes configurations, rotates credentials or suppresses alerts.
The second-order consequence is that identity, scoped permissions, evidence provenance and rollback mechanisms will become the practical rate-limiters on agentic defence. Organisations may be overinvesting in detection and orchestration while underbuilding the controls that make automated action auditable and safely revocable. This trend will be confirmed if future guidance converges on agent-specific authorization, action logs and recovery testing, rather than model performance alone.
#AgenticAI #CyberSecurity #IdentitySecurity #AIAssurance
NIST’s warning that early agentic deployments are prioritising features and immediate value over security identifies the core architectural gap: agents require an identity foundation before they can be trusted with consequential enterprise actions. At the same time, NIST is seeking input on AI-enabled vulnerability management for the NVD, including automation opportunities in a field where data drives compliance and remediation. The implication is significant. AI may accelerate the conversion of vulnerability intelligence into action, but the risk is no longer merely an inaccurate finding. It is an agent acting on inaccurate, incomplete or adversarially influenced evidence.
The NCSC’s observation that defenders cannot use AI in the same way as attackers sharpens this asymmetry. Attackers need only exploit an opening; defenders must establish authority, preserve availability and justify intervention. Recent NCSC calls for immediate action on PHP and actively exploited Citrix NetScaler vulnerabilities show why pressure to automate will intensify. Yet urgency cannot substitute for accountable execution.
Conventional AI-security thinking still treats autonomy as a graduated capability problem: begin with a copilot, then permit more actions. The more important distinction is between reversible assistance and authority-bearing operations. An agent that drafts a remediation plan is fundamentally different from one that changes configurations, rotates credentials or suppresses alerts.
The second-order consequence is that identity, scoped permissions, evidence provenance and rollback mechanisms will become the practical rate-limiters on agentic defence. Organisations may be overinvesting in detection and orchestration while underbuilding the controls that make automated action auditable and safely revocable. This trend will be confirmed if future guidance converges on agent-specific authorization, action logs and recovery testing, rather than model performance alone.
#AgenticAI #CyberSecurity #IdentitySecurity #AIAssurance
0
0
2
51
Oil prices hovers near Iran high amid Mideast supply jitters
investing.com/news/commoditi…
#OilPrices #IranTensions #MiddleEastCrisis #EnergyMarkets #CrudeOil
$CL $BZ $USO $XLE $OXY
investing.com/news/commoditi…
#OilPrices #IranTensions #MiddleEastCrisis #EnergyMarkets #CrudeOil
$CL $BZ $USO $XLE $OXY
0
0
0
39
Gold set for steepest weekly drop in over four months as prices fall below $4,000
investing.com/news/commoditi…
#Gold #XAUUSD #FedRates #Commodities #SafeHaven
$GLD $GOLD $IAU $NEM $GDX
investing.com/news/commoditi…
#Gold #XAUUSD #FedRates #Commodities #SafeHaven
$GLD $GOLD $IAU $NEM $GDX
0
0
1
50
Analysis-Trump’s diesel price relief efforts grow costlier as election nears
investing.com/news/commoditi…
#DieselPrices #Trump #USElection #EnergyMarkets #Inflation
$USO $XLE $CVX $XOM $OXY
investing.com/news/commoditi…
#DieselPrices #Trump #USElection #EnergyMarkets #Inflation
$USO $XLE $CVX $XOM $OXY
0
0
0
14
Oil prices jump $3 on Houthi attacks, oil spill, highest settlement since April
investing.com/news/commoditi…
#OilPrices #CrudeOil #HouthiAttacks #EnergyMarkets #WTI $CL $USO $XLE
investing.com/news/commoditi…
#OilPrices #CrudeOil #HouthiAttacks #EnergyMarkets #WTI $CL $USO $XLE
0
0
0
117
India’s crude oil imports climb 16% in September
investing.com/news/commoditi…
#IndiaOilImports #CrudeOil #EnergyMarkets #OilDemand #IndiaEconomy
$CL $USO $XOM $CVX $IOC
investing.com/news/commoditi…
#IndiaOilImports #CrudeOil #EnergyMarkets #OilDemand #IndiaEconomy
$CL $USO $XOM $CVX $IOC
0
0
1
88
Silver tests $53.08 support in oversold downtrend: Live levels
investing.com/news/commoditi…
#Silver #XAGUSD #PreciousMetals #Commodities #TechnicalAnalysis
$SLV $SIVR $AG
investing.com/news/commoditi…
#Silver #XAGUSD #PreciousMetals #Commodities #TechnicalAnalysis
$SLV $SIVR $AG
0
0
0
99
AI security assurance is converging on a more prosaic bottleneck: not whether an agent can reason, but whether every machine action can be tied to a verifiable identity, narrowly delegated authority and trustworthy security evidence.
NIST’s warning that early agentic deployments are prioritising features and ROI over identity foundations is therefore more consequential than it first appears. As agents move into purchasing, customer service, security and software development, identity is no longer merely an access-control concern. It becomes the mechanism that distinguishes an authorised automated action from an untraceable chain of tool calls, data retrievals and downstream changes.
The NCSC’s observation that defenders cannot use AI in the same way attackers can sharpens the point. Attackers can exploit scale with comparatively few constraints; defenders must operate inside production environments where an incorrect automated action can interrupt services, expose data or impair recovery. The result is an emerging asymmetry: defensive agent capability will be bounded less by model performance than by the quality of permissions, logging, rollback paths and evidence used to justify action.
This helps explain why NIST is seeking input on AI-enabled vulnerability management while the NCSC is examining external attack-surface management. AI can accelerate discovery and prioritisation, but it also turns vulnerability intelligence into an increasingly automated control-plane input. If evidence quality, provenance and asset identity are weak, faster analysis may simply produce faster misprioritisation or unsafe remediation.
Organisations may be treating agent assurance as a testing problem: evaluate the model, then govern deployment. The stronger pattern is that assurance must be designed into the operational substrate before autonomy expands. The decisive evidence will be whether future agentic-security guidance specifies machine-verifiable delegation, action-level audit trails and containment requirements, rather than primarily better detection or evaluation methods.
#AgenticAI #IdentitySecurity #AIAssurance #CyberResilience
NIST’s warning that early agentic deployments are prioritising features and ROI over identity foundations is therefore more consequential than it first appears. As agents move into purchasing, customer service, security and software development, identity is no longer merely an access-control concern. It becomes the mechanism that distinguishes an authorised automated action from an untraceable chain of tool calls, data retrievals and downstream changes.
The NCSC’s observation that defenders cannot use AI in the same way attackers can sharpens the point. Attackers can exploit scale with comparatively few constraints; defenders must operate inside production environments where an incorrect automated action can interrupt services, expose data or impair recovery. The result is an emerging asymmetry: defensive agent capability will be bounded less by model performance than by the quality of permissions, logging, rollback paths and evidence used to justify action.
This helps explain why NIST is seeking input on AI-enabled vulnerability management while the NCSC is examining external attack-surface management. AI can accelerate discovery and prioritisation, but it also turns vulnerability intelligence into an increasingly automated control-plane input. If evidence quality, provenance and asset identity are weak, faster analysis may simply produce faster misprioritisation or unsafe remediation.
Organisations may be treating agent assurance as a testing problem: evaluate the model, then govern deployment. The stronger pattern is that assurance must be designed into the operational substrate before autonomy expands. The decisive evidence will be whether future agentic-security guidance specifies machine-verifiable delegation, action-level audit trails and containment requirements, rather than primarily better detection or evaluation methods.
#AgenticAI #IdentitySecurity #AIAssurance #CyberResilience
0
0
0
21
Natural Gas hits oversold $2.633, tests key support: Live levels
investing.com/news/commoditi…
#NaturalGas #NatGas #Commodities #EnergyMarkets #TechnicalAnalysis
$NG $UNG $BOIL $KOLD
investing.com/news/commoditi…
#NaturalGas #NatGas #Commodities #EnergyMarkets #TechnicalAnalysis
$NG $UNG $BOIL $KOLD
0
0
0
284
Brent Oil hits $117.10 overbought RSI 74: Live levels
investing.com/news/commoditi…
#BrentCrude #OilPrices #EnergyMarkets #Commodities #RSI
$BZ=F $CL=F $XOM $CVX $OXY
investing.com/news/commoditi…
#BrentCrude #OilPrices #EnergyMarkets #Commodities #RSI
$BZ=F $CL=F $XOM $CVX $OXY
0
0
0
23
AI assurance is shifting from provider claims and benchmark scores toward independently observable evidence of what AI systems actually do in operation.
The strongest signal is the search for assurance surfaces that do not require trusting the operator. Research on detecting AI training through datacentre power proposes passive, facility-level measurement to distinguish training from declared inference workloads. Its significance is not the reported detection rate alone, but the governance model: compute commitments could be checked through physical side effects rather than self-reported logs.
The same logic appears inside AI applications. A new survey of 118 RAG-protection studies finds that defenses are frequently evaluated against incompatible endpoints, while its testbed shows that blocking verbatim copying did not prevent adaptive recovery of hidden facts. Assurance cannot rest on a control’s stated mechanism, such as reranking or context rewriting. It must test the security outcome an adaptive adversary can achieve.
This is also reshaping defensive automation. SESAA proposes LLM-generated static-analysis rules that remain executable and human-readable, creating an inspectable artifact rather than an opaque detection recommendation. NIST’s warning that agentic AI needs a strong identity foundation completes the picture: an agent’s actions must be attributable to a specific, bounded authority before its outputs, tool calls, or security decisions can become credible evidence.
What organisations may be getting wrong is treating audit logs, model evaluations, and safety filters as assurance by default. Those are assertions generated within the system being assessed. The emerging architecture separates actor, control, and verifier: identity constrains action; executable policies express the control; and independent observation tests the outcome.
The second-order implication is that the most valuable AI-security telemetry may sit outside the model stack, in permissions, tool execution, retrieval traces, physical infrastructure, and adversarial outcome testing. This trend will be confirmed if procurement and regulation begin demanding independently reproducible evidence, not merely vendor attestations or aggregate benchmark results.
#AIAssurance #AIsecurity #AgenticAI #Governance
The strongest signal is the search for assurance surfaces that do not require trusting the operator. Research on detecting AI training through datacentre power proposes passive, facility-level measurement to distinguish training from declared inference workloads. Its significance is not the reported detection rate alone, but the governance model: compute commitments could be checked through physical side effects rather than self-reported logs.
The same logic appears inside AI applications. A new survey of 118 RAG-protection studies finds that defenses are frequently evaluated against incompatible endpoints, while its testbed shows that blocking verbatim copying did not prevent adaptive recovery of hidden facts. Assurance cannot rest on a control’s stated mechanism, such as reranking or context rewriting. It must test the security outcome an adaptive adversary can achieve.
This is also reshaping defensive automation. SESAA proposes LLM-generated static-analysis rules that remain executable and human-readable, creating an inspectable artifact rather than an opaque detection recommendation. NIST’s warning that agentic AI needs a strong identity foundation completes the picture: an agent’s actions must be attributable to a specific, bounded authority before its outputs, tool calls, or security decisions can become credible evidence.
What organisations may be getting wrong is treating audit logs, model evaluations, and safety filters as assurance by default. Those are assertions generated within the system being assessed. The emerging architecture separates actor, control, and verifier: identity constrains action; executable policies express the control; and independent observation tests the outcome.
The second-order implication is that the most valuable AI-security telemetry may sit outside the model stack, in permissions, tool execution, retrieval traces, physical infrastructure, and adversarial outcome testing. This trend will be confirmed if procurement and regulation begin demanding independently reproducible evidence, not merely vendor attestations or aggregate benchmark results.
#AIAssurance #AIsecurity #AgenticAI #Governance
1
0
1
40